The document summarizes research on developing an AI actor for an online improvisational drama system. The AI actor has an affect detection component to analyze text input from human players and respond appropriately. Key aspects discussed include:
1) The affect detection analyzes linguistic cues like imperative mood, pattern matching, and semantic analysis to interpret emotion.
2) Metaphorical language is common in the drama and the system analyzes metaphors representing emotions as physical entities or food terms.
3) Context profiles allow the system to interpret affect based on background information about characters and scenarios.
4) An evaluation of the AI actor found it helped stimulate the improvisation in various situations.
State-of-the-Art Text Classification using Deep Contextual Word RepresentationsAusaf Ahmed
Introduced a state-of-the-art text classifier by addressing the capability of language semantics and polysemy in Natural Language Processing tasks. Used contextual representations of a word to achieve a ~5% increase in metrics outperforming existing models.
Several attempts had been made to analyze emotion words in the fields of linguistics, psychology and sociology; with the advent of computers, the analyses of these words have taken a different dimension. Unfortunately, limited attempts have so far been made to using interval type-2 fuzzy logic (IT2FL) to analyze these words in native languages. This study used IT2FL to analyze Igbo emotion words. IT2F sets are computed using the interval approach method which is divided into two parts: the data part and the fuzzy set part. The data part preprocessed data and its statistics computed for the interval that survived the preprocessing stages while the fuzzy set part determined the nature of the footprint of uncertainty; the IT2F set mathematical models for each emotion characteristics of each emotion word is also computed. The data used in this work was collected from fifteen subjects who were asked to enter an interval for each of the emotion characteristics: Valence, Activation and Dominance on an interval survey of the thirty Igbo emotion words. With this, the words are being analyzed and can be used for the purposes of translation between vocabularies in consideration to context.
PRONOUN DISAMBIGUATION: WITH APPLICATION TO THE WINOGRAD SCHEMA CHALLENGEkevig
A value-based approach to Natural Language Understanding, in particular, the disambiguation of
pronouns, is illustrated with a solution to a typical example from the Winograd Schema Challenge. The
worked example uses a language engine, Enguage, to support the articulation of the advocation and
fearing of violence. The example illustrates the indexical nature of pronouns, and how their values, their
referent objects, change because they are set by contextual data. It must be noted that Enguage is not a
suitable candidate for addressing the Winograd Schema Challenge as it is an interactive tool, whereas
the Challenge requires a preconfigured, unattended program.
OPTIMIZE THE LEARNING RATE OF NEURAL ARCHITECTURE IN MYANMAR STEMMERijnlc
Morphological stemming becomes a critical step toward natural language processing. The process of stemming is to reduce alternative forms to a common morphological root. Word segmentation for Myanmar Language, like for most Asian Languages, is an important task and extensively-studied sequence labelling problem. Named entity detection is one of the issues in Asian Language that has traditionally required a large amount of feature engineering to achieve high performance. The new approach is integrating them that would benefit in all these processes. In recent years, end-to-end sequence labelling models with deep learning are widely used. This paper introduces a deep BiGRUCNN-CRF network that jointly learns word segmentation, stemming and named entity recognition tasks. We trained the model using manually annotated corpora. State-of-the-art named entity recognition systems rely heavily on handcrafted feature built in our new approach, we introduce the joint model that relies on two sources of information: character level representation and syllable level representation.
State-of-the-Art Text Classification using Deep Contextual Word RepresentationsAusaf Ahmed
Introduced a state-of-the-art text classifier by addressing the capability of language semantics and polysemy in Natural Language Processing tasks. Used contextual representations of a word to achieve a ~5% increase in metrics outperforming existing models.
Several attempts had been made to analyze emotion words in the fields of linguistics, psychology and sociology; with the advent of computers, the analyses of these words have taken a different dimension. Unfortunately, limited attempts have so far been made to using interval type-2 fuzzy logic (IT2FL) to analyze these words in native languages. This study used IT2FL to analyze Igbo emotion words. IT2F sets are computed using the interval approach method which is divided into two parts: the data part and the fuzzy set part. The data part preprocessed data and its statistics computed for the interval that survived the preprocessing stages while the fuzzy set part determined the nature of the footprint of uncertainty; the IT2F set mathematical models for each emotion characteristics of each emotion word is also computed. The data used in this work was collected from fifteen subjects who were asked to enter an interval for each of the emotion characteristics: Valence, Activation and Dominance on an interval survey of the thirty Igbo emotion words. With this, the words are being analyzed and can be used for the purposes of translation between vocabularies in consideration to context.
PRONOUN DISAMBIGUATION: WITH APPLICATION TO THE WINOGRAD SCHEMA CHALLENGEkevig
A value-based approach to Natural Language Understanding, in particular, the disambiguation of
pronouns, is illustrated with a solution to a typical example from the Winograd Schema Challenge. The
worked example uses a language engine, Enguage, to support the articulation of the advocation and
fearing of violence. The example illustrates the indexical nature of pronouns, and how their values, their
referent objects, change because they are set by contextual data. It must be noted that Enguage is not a
suitable candidate for addressing the Winograd Schema Challenge as it is an interactive tool, whereas
the Challenge requires a preconfigured, unattended program.
OPTIMIZE THE LEARNING RATE OF NEURAL ARCHITECTURE IN MYANMAR STEMMERijnlc
Morphological stemming becomes a critical step toward natural language processing. The process of stemming is to reduce alternative forms to a common morphological root. Word segmentation for Myanmar Language, like for most Asian Languages, is an important task and extensively-studied sequence labelling problem. Named entity detection is one of the issues in Asian Language that has traditionally required a large amount of feature engineering to achieve high performance. The new approach is integrating them that would benefit in all these processes. In recent years, end-to-end sequence labelling models with deep learning are widely used. This paper introduces a deep BiGRUCNN-CRF network that jointly learns word segmentation, stemming and named entity recognition tasks. We trained the model using manually annotated corpora. State-of-the-art named entity recognition systems rely heavily on handcrafted feature built in our new approach, we introduce the joint model that relies on two sources of information: character level representation and syllable level representation.
Input processing (IP) is concerned with these situations, the reason being that acquisition is, to a certain degree, a by-product of comprehension (see, e.g., Truscott & Sharwood Smith, 2004)
A SURVEY OF S ENTIMENT CLASSIFICATION TECHNIQUES USED FOR I NDIAN REGIONA...ijcsa
Sentiment Analysis is a natural language processing
task that extracts sentiment from various text for
ms
and classifies them according to positive, negative
or neutral polarity. It analyzes emotions, feeling
s, and
the attitude of a speaker or a writer towards a con
text. This paper gives comparative study of various
sentiment classification techniques and also discus
ses in detail two main categories of sentiment
classification techniques these are machine based a
nd lexicon based. The paper also presents challenge
s
associated with sentiment analysis along with lexic
al resources available.
Words can have more than one distinct meaning and many words can be interpreted in multiple ways
depending on the context in which they occur. The process of automatically identifying the meaning of
a polysemous word in a sentence is a fundamental task in Natural Language Processing (NLP). This
phenomenon poses challenges to Natural Language Processing systems. There have been many efforts
on word sense disambiguation for English; however, the amount of efforts for Amharic is very little.
Many natural language processing applications, such as Machine Translation, Information Retrieval,
Question Answering, and Information Extraction, require this task, which occurs at the semantic level.
In this thesis, a knowledge-based word sense disambiguation method that employs Amharic WordNet
is developed. Knowledge-based Amharic WSD extracts knowledge from word definitions and relations
among words and senses. The proposed system consists of preprocessing, morphological analysis and
disambiguation components besides Amharic WordNet database. Preprocessing is used to prepare the
input sentence for morphological analysis and morphological analysis is used to reduce various forms
of a word to a single root or stem word. Amharic WordNet contains words along with its different
meanings, synsets and semantic relations with in concepts. Finally, the disambiguation component is
used to identify the ambiguous words and assign the appropriate sense of ambiguous words in a
sentence using Amharic WordNet by using sense overlap and related words.
We have evaluated the knowledge-based Amharic word sense disambiguation using Amharic
WordNet system by conducting two experiments. The first one is evaluating the effect of Amharic
WordNet with and without morphological analyzer and the second one is determining an optimal
windows size for Amharic WSD. For Amharic WordNet with morphological analyzer and Amharic
WordNet without morphological analyzer we have achieved an accuracy of 57.5% and 80%,
respectively. In the second experiment, we have found that two-word window on each side of the
ambiguous word is enough for Amharic WSD. The test results have shown that the proposed WSD
methods have performed better than previous Amharic WSD methods.
Keywords: Natural Language Processing, Amharic WordNet, Word Sense Disambiguation,
Knowledge Based Approach, Lesk Algorithm
Psychology of Language 5th Edition Carroll Test BankKiayadare
Full download : http://alibabadownload.com/product/psychology-of-language-5th-edition-carroll-test-bank/ Psychology of Language 5th Edition Carroll Test Bank
Presentación2.ppt input and interactionJoel Acosta
The primary factor affecting language acquisition appears to be the input that the learner receives. Stephen Krashen took a very strong position on the importance of input, asserting that comprehensible input is all that is necessary for second-language acquisition.
Conversational interaction in a second language forms the basis for the development of language rather than being only a forum for practice of specific language features. (Gass, 2003)
Sentence level sentiment polarity calculation for customer reviews by conside...eSAT Publishing House
IJRET : International Journal of Research in Engineering and Technology is an international peer reviewed, online journal published by eSAT Publishing House for the enhancement of research in various disciplines of Engineering and Technology. The aim and scope of the journal is to provide an academic medium and an important reference for the advancement and dissemination of research results that support high-level learning, teaching and research in the fields of Engineering and Technology. We bring together Scientists, Academician, Field Engineers, Scholars and Students of related fields of Engineering and Technology
On Tracking Behavior of Streaming Data: An Unsupervised ApproachWaqas Tariq
In the recent years, data streams have been in the gravity of focus of quite a lot number of researchers in different domains. All these researchers share the same difficulty when discovering unknown pattern within data streams that is concept change. The notion of concept change refers to the places where underlying distribution of data changes from time to time. There have been proposed different methods to detect changes in the data stream but most of them are based on an unrealistic assumption of having data labels available to the learning algorithms. Nonetheless, in the real world problems labels of streaming data are rarely available. This is the main reason why data stream communities have recently focused on unsupervised domain. This study is based on the observation that unsupervised approaches for learning data stream are not yet matured; namely, they merely provide mediocre performance specially when applied on multi-dimensional data streams. In this paper, we propose a method for Tracking Changes in the behavior of instances using Cumulative Density Function; abbreviated as TrackChCDF. Our method is able to detect change points along unlabeled data stream accurately and also is able to determine the trend of data called closing or opening. The advantages of our approach are three folds. First, it is able to detect change points accurately. Second, it works well in multi-dimensional data stream, and the last but not the least, it can determine the type of change, namely closing or opening of instances over the time which has vast applications in different fields such as economy, stock market, and medical diagnosis. We compare our algorithm to the state-of-the-art method for concept change detection in data streams and the obtained results are very promising.
Linking Design, Marketing, and Innovation: Managing the Connection for Compet...Waqas Tariq
Marketing should closely coordinate with product innovation. Not only will an effective connection of the marketing activity and design trigger product innovation in enterprises, it is a key factor to the product success in the market place. In practice, how to process product design so as to reach the goal of marketing during the innovation procedure? In the past, enterprises did have some concrete actions but empirical studies regarding the correlation model between marketing strategy and design strategy or even between marketing strategy and innovation are rarely seen. In the study, the author first synthesized viewpoints from literature on marketing strategy, product innovation strategy, and design strategy. Then managers at marketing divisions and R&D departments of the Taiwanese consumer electronic industry were interviewed. The fitness of theoretical models and the observed data was then verified by Structural Equation Modeling (SEM). From the 265 effective questionnaires, a proper fitness was found for the correlation theoretical model of marketing strategy, product innovation strategy, and design strategy, and the observed data. In enterprises, marketing strategy will influence design strategy, and at the same time, the marketing strategy will also influence design strategy through innovation strategy. Among these three variables, innovation strategy serves as an independent variable and intervene variable to design strategy.
Comparative Calibration Method Between two Different Wavelengths With Aureole...Waqas Tariq
A multi-stage method for calibration of sunphotometer is proposed by combining comparison calibration method between two different wavelengths with aureole observation method for long wavelength calibration. Its effectiveness in reducing the influences for calibration due to molecular and aerosolfs extinction in the unstable turbidity conditions is clarified. By comparing the calculated results with the proposed method and the existing individually calibration method, it is found that the proposed method is superior to the existing method in terms of calibration accuracy. Namely, Through a comparison between ILM and the proposed method using band 0.87um as reference, the largest calibration errors are 0.0014, 0.0428 by PM are lower than that by ILM (0.011,0.0489) for sky radiances with no error and -3~+3%, -5~+5% errors. By analyzing the observation data of 15 days with POM-1 Skyradiometer, the largest standard deviation of calibration constants by PM is 0.02016, and is lower than that by ILM (0.03858).
Determination of the Operational Parameters of a Planar Robot with Three JointsWaqas Tariq
Robots are currently made in numerous types and are used in diverse roles such as production lines, daily living activities and some security fields. These types of robots are well designed and successfully applied in many areas requiring high sensitivity and stability. The aim of this study was to determine the optimum values of several operational parameters for a planar robot with respect to robot design and construction. With this aim, a small planar robot with a three-jointed arm activated by hydraulic cylinders in each segment was evaluated using a technical design drawing. The arm motions of the planar robot are rotary and parallel within a vertical plane. The resulting optimal operational parameters of the planar robot were determined as starting and target positions of 31.5 cm and 55 cm, respectively, on the x-axis and 17.18 cm and 118.44 cm on the y–axis. Time-position and time-velocity graphs were constructed corresponding to the orbit-planning parameters, resulting in Cartesian velocities for the terminal processor of 13.98 m/sec on the x-axis and 20.16 m/sec on the y-axis at 1.5 seconds after initiation. The maximum power consumption of the robot was determined as 1 kW according to the outer load and arm weights.
HRI for Interactive Humanoid Head Amir-II for Visual Tracking and Servoing of...Waqas Tariq
In this paper, we describe the HRI (Human-Robot Interaction) system developed to operate a humanoid robot head capable of visual tracking and servoing of human face through image processing. The robotic humanoid head named Amir-II, equipped with a camera and servoing mechanism is used as the platform. The Amir-II tracks the human face within the field-of-vision (FOV) while the servoing mechanism ensures the detected human face remains at the center of its FOV. The algorithm developed in this research utilizes the capability offered by scientific computing program MATLAB along with its Image Processing Toolbox. The algorithm basically compares the locations of the face in the image plane that is detected from the static face image captured from real-time video stream. The calculated difference is then used to produce appropriate motion command for the servo mechanism to keep track of the human face moving within the range of its FOV.
Input processing (IP) is concerned with these situations, the reason being that acquisition is, to a certain degree, a by-product of comprehension (see, e.g., Truscott & Sharwood Smith, 2004)
A SURVEY OF S ENTIMENT CLASSIFICATION TECHNIQUES USED FOR I NDIAN REGIONA...ijcsa
Sentiment Analysis is a natural language processing
task that extracts sentiment from various text for
ms
and classifies them according to positive, negative
or neutral polarity. It analyzes emotions, feeling
s, and
the attitude of a speaker or a writer towards a con
text. This paper gives comparative study of various
sentiment classification techniques and also discus
ses in detail two main categories of sentiment
classification techniques these are machine based a
nd lexicon based. The paper also presents challenge
s
associated with sentiment analysis along with lexic
al resources available.
Words can have more than one distinct meaning and many words can be interpreted in multiple ways
depending on the context in which they occur. The process of automatically identifying the meaning of
a polysemous word in a sentence is a fundamental task in Natural Language Processing (NLP). This
phenomenon poses challenges to Natural Language Processing systems. There have been many efforts
on word sense disambiguation for English; however, the amount of efforts for Amharic is very little.
Many natural language processing applications, such as Machine Translation, Information Retrieval,
Question Answering, and Information Extraction, require this task, which occurs at the semantic level.
In this thesis, a knowledge-based word sense disambiguation method that employs Amharic WordNet
is developed. Knowledge-based Amharic WSD extracts knowledge from word definitions and relations
among words and senses. The proposed system consists of preprocessing, morphological analysis and
disambiguation components besides Amharic WordNet database. Preprocessing is used to prepare the
input sentence for morphological analysis and morphological analysis is used to reduce various forms
of a word to a single root or stem word. Amharic WordNet contains words along with its different
meanings, synsets and semantic relations with in concepts. Finally, the disambiguation component is
used to identify the ambiguous words and assign the appropriate sense of ambiguous words in a
sentence using Amharic WordNet by using sense overlap and related words.
We have evaluated the knowledge-based Amharic word sense disambiguation using Amharic
WordNet system by conducting two experiments. The first one is evaluating the effect of Amharic
WordNet with and without morphological analyzer and the second one is determining an optimal
windows size for Amharic WSD. For Amharic WordNet with morphological analyzer and Amharic
WordNet without morphological analyzer we have achieved an accuracy of 57.5% and 80%,
respectively. In the second experiment, we have found that two-word window on each side of the
ambiguous word is enough for Amharic WSD. The test results have shown that the proposed WSD
methods have performed better than previous Amharic WSD methods.
Keywords: Natural Language Processing, Amharic WordNet, Word Sense Disambiguation,
Knowledge Based Approach, Lesk Algorithm
Psychology of Language 5th Edition Carroll Test BankKiayadare
Full download : http://alibabadownload.com/product/psychology-of-language-5th-edition-carroll-test-bank/ Psychology of Language 5th Edition Carroll Test Bank
Presentación2.ppt input and interactionJoel Acosta
The primary factor affecting language acquisition appears to be the input that the learner receives. Stephen Krashen took a very strong position on the importance of input, asserting that comprehensible input is all that is necessary for second-language acquisition.
Conversational interaction in a second language forms the basis for the development of language rather than being only a forum for practice of specific language features. (Gass, 2003)
Sentence level sentiment polarity calculation for customer reviews by conside...eSAT Publishing House
IJRET : International Journal of Research in Engineering and Technology is an international peer reviewed, online journal published by eSAT Publishing House for the enhancement of research in various disciplines of Engineering and Technology. The aim and scope of the journal is to provide an academic medium and an important reference for the advancement and dissemination of research results that support high-level learning, teaching and research in the fields of Engineering and Technology. We bring together Scientists, Academician, Field Engineers, Scholars and Students of related fields of Engineering and Technology
On Tracking Behavior of Streaming Data: An Unsupervised ApproachWaqas Tariq
In the recent years, data streams have been in the gravity of focus of quite a lot number of researchers in different domains. All these researchers share the same difficulty when discovering unknown pattern within data streams that is concept change. The notion of concept change refers to the places where underlying distribution of data changes from time to time. There have been proposed different methods to detect changes in the data stream but most of them are based on an unrealistic assumption of having data labels available to the learning algorithms. Nonetheless, in the real world problems labels of streaming data are rarely available. This is the main reason why data stream communities have recently focused on unsupervised domain. This study is based on the observation that unsupervised approaches for learning data stream are not yet matured; namely, they merely provide mediocre performance specially when applied on multi-dimensional data streams. In this paper, we propose a method for Tracking Changes in the behavior of instances using Cumulative Density Function; abbreviated as TrackChCDF. Our method is able to detect change points along unlabeled data stream accurately and also is able to determine the trend of data called closing or opening. The advantages of our approach are three folds. First, it is able to detect change points accurately. Second, it works well in multi-dimensional data stream, and the last but not the least, it can determine the type of change, namely closing or opening of instances over the time which has vast applications in different fields such as economy, stock market, and medical diagnosis. We compare our algorithm to the state-of-the-art method for concept change detection in data streams and the obtained results are very promising.
Linking Design, Marketing, and Innovation: Managing the Connection for Compet...Waqas Tariq
Marketing should closely coordinate with product innovation. Not only will an effective connection of the marketing activity and design trigger product innovation in enterprises, it is a key factor to the product success in the market place. In practice, how to process product design so as to reach the goal of marketing during the innovation procedure? In the past, enterprises did have some concrete actions but empirical studies regarding the correlation model between marketing strategy and design strategy or even between marketing strategy and innovation are rarely seen. In the study, the author first synthesized viewpoints from literature on marketing strategy, product innovation strategy, and design strategy. Then managers at marketing divisions and R&D departments of the Taiwanese consumer electronic industry were interviewed. The fitness of theoretical models and the observed data was then verified by Structural Equation Modeling (SEM). From the 265 effective questionnaires, a proper fitness was found for the correlation theoretical model of marketing strategy, product innovation strategy, and design strategy, and the observed data. In enterprises, marketing strategy will influence design strategy, and at the same time, the marketing strategy will also influence design strategy through innovation strategy. Among these three variables, innovation strategy serves as an independent variable and intervene variable to design strategy.
Comparative Calibration Method Between two Different Wavelengths With Aureole...Waqas Tariq
A multi-stage method for calibration of sunphotometer is proposed by combining comparison calibration method between two different wavelengths with aureole observation method for long wavelength calibration. Its effectiveness in reducing the influences for calibration due to molecular and aerosolfs extinction in the unstable turbidity conditions is clarified. By comparing the calculated results with the proposed method and the existing individually calibration method, it is found that the proposed method is superior to the existing method in terms of calibration accuracy. Namely, Through a comparison between ILM and the proposed method using band 0.87um as reference, the largest calibration errors are 0.0014, 0.0428 by PM are lower than that by ILM (0.011,0.0489) for sky radiances with no error and -3~+3%, -5~+5% errors. By analyzing the observation data of 15 days with POM-1 Skyradiometer, the largest standard deviation of calibration constants by PM is 0.02016, and is lower than that by ILM (0.03858).
Determination of the Operational Parameters of a Planar Robot with Three JointsWaqas Tariq
Robots are currently made in numerous types and are used in diverse roles such as production lines, daily living activities and some security fields. These types of robots are well designed and successfully applied in many areas requiring high sensitivity and stability. The aim of this study was to determine the optimum values of several operational parameters for a planar robot with respect to robot design and construction. With this aim, a small planar robot with a three-jointed arm activated by hydraulic cylinders in each segment was evaluated using a technical design drawing. The arm motions of the planar robot are rotary and parallel within a vertical plane. The resulting optimal operational parameters of the planar robot were determined as starting and target positions of 31.5 cm and 55 cm, respectively, on the x-axis and 17.18 cm and 118.44 cm on the y–axis. Time-position and time-velocity graphs were constructed corresponding to the orbit-planning parameters, resulting in Cartesian velocities for the terminal processor of 13.98 m/sec on the x-axis and 20.16 m/sec on the y-axis at 1.5 seconds after initiation. The maximum power consumption of the robot was determined as 1 kW according to the outer load and arm weights.
HRI for Interactive Humanoid Head Amir-II for Visual Tracking and Servoing of...Waqas Tariq
In this paper, we describe the HRI (Human-Robot Interaction) system developed to operate a humanoid robot head capable of visual tracking and servoing of human face through image processing. The robotic humanoid head named Amir-II, equipped with a camera and servoing mechanism is used as the platform. The Amir-II tracks the human face within the field-of-vision (FOV) while the servoing mechanism ensures the detected human face remains at the center of its FOV. The algorithm developed in this research utilizes the capability offered by scientific computing program MATLAB along with its Image Processing Toolbox. The algorithm basically compares the locations of the face in the image plane that is detected from the static face image captured from real-time video stream. The calculated difference is then used to produce appropriate motion command for the servo mechanism to keep track of the human face moving within the range of its FOV.
Leadership Driven Innovation: The Role of the Engineer in Our FutureWaqas Tariq
The development of new technologies, particularly those identified as advanced and/or disruptive, rely on two individual but highly interrelated competencies, leadership and innovation. These two are the basis for the successful development of most of the major technologies in production, today. At their best they are also the genesis for most of the large commercial and industrial organizations currently operating in the global marketplace. More importantly, it is the state of health of these two competencies that determines the longevity and profitability of these organizations. This paper tracks a hypothetical progression from inception to integrated production for what can be idealized as a maturation process of a new company/technology. While it is directed to represent most any type of new technology it is particularly suited to advance technologies and those organizations dealing in products for accelerating markets, such as for the aerospace and advance communications arenas.
Integrating Threat Modeling in Secure Agent-Oriented Software DevelopmentWaqas Tariq
The main objective of this paper is to integrate threat modeling when developing a software application following the Secure Tropos methodology. Secure Tropos is an agent-oriented software development methodology which integrates “security extensions” into all development phases. Threat modeling is used to identify, document, and mitigate security risks, therefore, applying threat modeling when defining the security extensions shall lead to better modeling and increased level of security. After integrating threat modeling into this methodology, security attack scenarios are applied to the models to discuss how the security level of the system has been impacted. Security attack scenarios have been used to test different enhancements made to the Secure Tropos methodology and the Tropos methodology itself. The system modeled using this methodology is an e-Commerce application that will be used to sell handmade products made in Ecuador through the web. The .NET Model-View-Controller framework is used to develop our case study application. Results show that integrating threat modeling in the development process, the level of security of the modeled application has increased. The different actors, goals, tasks, and security constraints that were introduced based on the proposed integration help mitigate different risks and vulnerabilities.
Correlation of Shallow Groundwater Levels with The Liquefaction Occurrence Ca...Waqas Tariq
When a large earthquake hit Yogyakarta in 26 May 2006, several liquefaction events occurred in some places in Yogyakarta and surrounding areas. Liquefaction event is strongly influenced by the depth of the ground water in the area, as well as several other parameters. This paper will conduct a qualitative correlation between the observational data liquefaction after the earthquake and the measured groundwater depth.
A total of 493 water-table depths were directly measured in the southern part area of volcanic- clastic sediment by using a measuring-tape meter. Fairly high correlation is shown between areas with shallow water depth and the position of occurred liquefactions. The average water depth for liquefied soil is 2.05 m. Almost 90% of the study area showed a high potential for liquefaction to occur.
Optimum Algorithm for Computing the Standardized Moments Using MATLAB 7.10(R2...Waqas Tariq
A fundamental task in many statistical analyses is to characterize the location and variability of a data set. A further characterization of the data includes skewness and kurtosis. This paper emphasizes the real time computational problem for generally the rth standardized moments and specially for both skewness and kurtosis. It has therefore been important to derive an optimum computational technique for the standardized moments. A new algorithm has been designed for the evaluation of the standardized moments. The evaluation of error analysis has been discussed. The new algorithm saved computational energy by approximately 99.95% than that of the previously published algorithms.
Integrating Web Services With Geospatial Data Mining Disaster Management for ...Waqas Tariq
Data Mining (DM) and Geographical Information Systems (GIS) are complementary techniques for describing, transforming, analyzing and modeling data about real world system. GIS and DM are naturally synergistic technologies that can be joined to produce powerful market insight from a sea of disparate data. Web Services would greatly simplify the development of many kinds of data integration and knowledge management applications. This research aims to develop a Spatial DM web service. It integrates state of the art GIS and DM functionality in an open, highly extensible, web-based architecture. The Interoperability of geospatial data previously focus just on data formats and standards. The recent popularity and adoption of Web Services has provided new means of interoperability for geospatial information not just for exchanging data but for analyzing these data during exchange as well. An integrated, user friendly Spatial DM System available on the internet via a web service offers exciting new possibilities for geo-spatial analysis to be ready for decision making and geographical research to a wide range of potential users.
Movement of Share Prices and Sectoral Analysis: A Reflection Through Interact...Waqas Tariq
Interaction in graphs gives the user with an advantage to analyze the data in greater depth. With the help of interactive graphics users can get better insight of the data in comparison to the static graphical tools. This paper introduces an interactive graphical tool consisting of two graphs, a line diagram complemented by a boxplot. The line diagram helps to understand how successive values of a variable are related to time and box plot can help the visual comparison of several such variables. Here the line diagram is used to visualize share prices of a company corresponding to a number of days and the boxplot displays the position of the Share price of all companies in a particular sector. An investor in share market needs to consider a number of factors before making any decision about investment. Some of the factors influencing the decision are the performance of the particular security in recent past, its position in terms of share price in its own sector. The graphical technique used in this software tool shall be helpful while making investment decision.
Creative Accounting and Impact on Management Decision MakingWaqas Tariq
The study was conducted to appraise the impact of creative accounting on management decisions of selected companies listed in the Nigerian Stock Exchange. With the background, the main objective of the study includes the examination of the extent to which macro-manipulation of financial statement affects management decisions; to examine the extent to which macro-manipulation of financial statement affects share price performance; and to determine the impact of misreported assets and liabilities as well as making recommendations to help remedy some of the problems. The research method used was descriptive and the primary data collected were summarized and tabulated. These were picked in line with the hypothesis variables of the study so as to determine their validity. It was observed that the application of creativity in financial statement reporting significantly affects the decision of management to recapitalize the firm upward or dispose of it reserves. The study concluded that creative accounting through macro-manipulation of financial statements affects a firm’s price and capital market performance. In view of the study, the researcher recommended that the application of creative accounting on management decision should be to avoid misreporting of assets and liabilities in their financial report, and that management decision towards creative accounting should be geared towards the relative advantage principle and good corporate governance which encourage challenges to current ways of thinking and not manipulating for self interest.
Periodic Review Model for Determining Inventory Policy for Aircraft Consumabl...Waqas Tariq
This research is conducted to develop inventory policy of aircraft consumable spare parts which are needed on aircraft maintenance activity . In this research, we used periodic review model to determine the optimal policy of aircraft spare parts inventory. By using the periodic review model, we find optimal period of inventory review and maximum level of inventory. The optimal decision is determined based on the minimum total cost. We have classified consumable spare parts using ABC method to categorize them based on their dollar contribution and demand frequency. Therefore in this research, we focus on managing the inventory level for spare parts on class C. The result from this study shows that the proposed periodic review policy result in lower total inventory cost compared the the company policy. The proposed policy gives an average saving 35.38 %.
Identification and Control of Three-Links Electrically Driven Robot Arm Using...Waqas Tariq
This paper uses a fuzzy neural network (FNN) structure for identifying and controlling nonlinear dynamic systems such three links robot arm. The equation of motion for three links robot arm derived using Lagrange’s equation. This equation then combined with the equations of motion for dc. servo motors which actuated the robot. For the control problem, we present the forward and inverse adaptive control approaches using the FNN. Computer simulation is performed to view the results for identification and control
Two Phase Algorithm for Solving VRPTW ProblemWaqas Tariq
Vehicle Routing Problem with Time Windows (VRPTW) is a well known NP hard combinatorial scheduling optimization problem in which minimum number of routes have to be determined to serve all the customers within their specified time windows. Different analytic and heuristic approaches have been tried to solve such problems. In this paper we propose a two phase method which utilizes Genetic algorithms as well as random search incorporating simulated annealing concepts to solve VRPTW problem in various scenarios.
Measurement and Comparison of Productivity Performance Under Fuzzy Imprecise ...Waqas Tariq
The creation of goods and services requires changing the expended resources into the output goods and services. How efficiently we transform these input resources into goods and services depends on the productivity of the transformation process. However, it has been observed there is always a vagueness or imprecision associated with the values of inputs and outputs. Therefore, it becomes hard for a productivity measurement expert to specify the amount of resources and the outputs as exact scalar numbers. The present paper, applies fuzzy set theory to measure and compare productivity performance of transformation processes when numerical data cannot be specified in exact terms. The approach makes it possible to measure and compare productivity of organizational units (including non-government and non-profit entities) when the expert inputs can not be specified as exact scalar quantities. The model has been applied to compare productivity of different branches of a company.
Value Creation Through Corporate Social Responsibility in Developing Countrie...Waqas Tariq
Consumer support for Corporate Social Responsibility (CSR) has been in practice for some years now and firms are demanded to seriously take CSR initiatives. This project has been compiled out of a wealth of literature that addresses the need and importance of CSR and business ethics in the society in great depth. A case study of CSR at Proctor and Gamble Pakistan was carried out by employing both the qualitative and quantitative data collection techniques to gather information so as to bring the attributes of triangulation in this research. The research findings outlined various views and beliefs of the respondents with regards to CSR initiatives by Proctor and Gamble Pakistan. The CSR awareness and societal veracities are the factors that encourage consumers to think ethically and make decisions in terms of who to develop associations with. The research shows a reflection of deductive approach and the researcher understood the inbuilt pros and cons of dependence upon secondary sources of information. It was attempted to adopt a hybrid strategy in this project but it mainly took a positivist look because of the nature of the questionnaire survey based upon close-ended questions aiming for quantitative data. The trend for CSR initiatives in developing countries is now growing at a decent pace and the recent advancements in technology and media have resulted in grown awareness among consumer groups to exert pressures on multinational companies to be apparent in their statements as well as practices.
Defect Management Practices and Problems in Free/Open Source Software ProjectsWaqas Tariq
With the advent of Free/Open Source Software (F/OSS) paradigm, a large number of projects are evolving that make use of F/OSS infrastructure and development practices. Defect Management System is an important component in F/OSS infrastructure which maintains defect records as well as tracks their status. The defect data comprising more than 60,000 defect reports from 20 F/OSS Projects is analyzed from various perspectives, with special focus on evaluating the efficiency and effectiveness in resolving defects and determining responsiveness towards users. Major problems and inefficiencies encountered in Defect Management among F/OSS Projects have been identified. A process is proposed to distribute roles and responsibilities among F/OSS participants which can help F/OSS Projects to improve the effectiveness and efficiency of Defect Management and hence assure better quality of F/OSS Projects.
Determining The Barriers Faced By Novice ProgrammersWaqas Tariq
Most of the novice programmers find glitches at various phases while trying to complete a program in their Computer Science programming course. These phases can be while constructing the code, finding errors in the code at the time of compilation of the program, debugging these errors while executing the program. Novice programmers are unable to understand some of the concepts in programming. Computer Science programming course instructors are experiencing difficulty in finding these barriers faced by the students. These barriers are forcing students to drop programming course from their degree plan and becoming a concern to the professors teaching programming course. In this research ActivePresenter software is used. This software recorded the full motion video with crystal clear quality and helped in capturing screen shots automatically with a click of a mouse or pressing any key in the keyboard of the students who are trying to complete a programming assignment. By analyzing all the recordings collected from different students, these barriers are determined.
Emotion Detection is one of the most emerging issues in human computer interaction. A sufficient amount
of work has been done by researchers to detect emotions from facial and audio information whereas
recognizing emotions from textual data is still a fresh and hot research area. This paper presented a
knowledge based survey on emotion detection based on textual data and the methods used for this purpose.
At the next step paper also proposed a new architecture for recognizing emotions from text document.
Proposed architecture is composed of two main parts, emotion ontology and emotion detector algorithm.
Proposed emotion detector system takes a text document and the emotion ontology as inputs and produces
one of the six emotion classes (i.e. love, joy, anger, sadness, fear and surprise) as the output.
BASIC ANALYSIS ON PROSODIC FEATURES IN EMOTIONAL SPEECHIJCSEA Journal
Speech is a rich source of information which gives not only about what a speaker says, but also about what the speaker’s attitude is toward the listener and toward the topic under discussion—as well as the speaker’s own current state of mind. Recently increasing attention has been directed to the study of the emotional content of speech signals, and hence, many systems have been proposed to identify the emotional content of a spoken utterance. The focus of this research work is to enhance man machine interface by focusing on user’s speech emotion. This paper gives the results of the basic analysis on prosodic features and also compares the prosodic features
of, various types and degrees of emotional expressions in Tamil speech based on the auditory impressions between the two genders of speakers as well as listeners. The speech samples consist of “neutral” speech as well as speech with three types of emotions (“anger”, “joy”, and “sadness”) of three degrees (“light”, “medium”, and “strong”). A listening test is also being conducted using 300 speech samples uttered by students at the ages of 19 -22 the ages of 19-22 years old. The features of prosodic parameters based on the emotional speech classified according to the auditory impressions of the subjects are analyzed. Analysis results suggest that prosodic features that identify their emotions and degrees are not only speakers’ gender dependent, but also listeners’ gender dependent.
USER EXPERIENCE AND DIGITALLY TRANSFORMED/CONVERTED EMOTIONSIJMIT JOURNAL
In human natural interaction (human-human interaction), humans use speech beside the non-verbal cues
like facial expressions movements and gesture movements to express themselves. However, in (humancomputer
interactions), computer will use the non-verbal cues of human beings to determine the user
experience and usability of any software or application on the computer. We introduce a new model called
Measuring User Experience using Digitally Transformed/Converted Emotions (MUDE) which measures
two metrics of user experience(satisfaction and errors) , and compares them with SUS questionnaire results
by conducting an experiment for measuring the usability and user’s experience.
Signal Processing Tool for Emotion Recognitionidescitation
In the course of realization of modern day robots,
which not only perform tasks, but also behaves like human
beings during their interaction with the natural environment,
it is essential for us to impart knowledge of the underlying
emotions in the spoken utterances of human beings to the
robots, enabling them to be consistent, whole, complete and
perfect. To this end, it is essential for them too to understand
and identify the human emotions. For this reason, stress is
laid now-a-days on the study of emotional content of the speech
and accordingly speech emotion recognition engines have been
proposed. This paper is a survey of the main aspects of speech
emotion recognition, namely, features extractions and types
of features commonly used, selection of most informed
features from the original dataset of the features, and
classification of the features according to different classifying
techniques based on relative information regarding commonly
used database for the speech emotion recognition.
Natural Language Processing: A comprehensive overviewBenjaminlapid1
Natural language processing enhances human-computer interaction by bridging the language gap. Uncover its applications and techniques in this comprehensive overview. Dive in now!
Natural Language Processing: State of The Art, Current Trends and Challengesantonellarose
Diksha Khurana1
, Aditya Koli1
, Kiran Khatter1,2 and Sukhdev Singh1,2
1Department of Computer Science and Engineering
Manav Rachna International University, Faridabad-121004, India
2Accendere Knowledge Management Services Pvt. Ltd., India
Similar to Affect Sensing and Contextual Affect Modeling from Improvisational Interaction (20)
The Use of Java Swing’s Components to Develop a WidgetWaqas Tariq
Widget is a kind of application provides a single service such as a map, news feed, simple clock, battery-life indicators, etc. This kind of interactive software object has been developed to facilitate user interface (UI) design. A user interface (UI) function may be implemented using different widgets with the same function. In this article, we present the widget as a platform that is generally used in various applications, such as in desktop, web browser, and mobile phone. We also describe a visual menu of Java Swing’s components that will be used to establish widget. It will assume that we have successfully compiled and run a program that uses Swing components.
3D Human Hand Posture Reconstruction Using a Single 2D ImageWaqas Tariq
Passive sensing of the 3D geometric posture of the human hand has been studied extensively over the past decade. However, these research efforts have been hampered by the computational complexity caused by inverse kinematics and 3D reconstruction. In this paper, our objective focuses on 3D hand posture estimation based on a single 2D image with aim of robotic applications. We introduce the human hand model with 27 degrees of freedom (DOFs) and analyze some of its constraints to reduce the DOFs without any significant degradation of performance. A novel algorithm to estimate the 3D hand posture from eight 2D projected feature points is proposed. Experimental results using real images confirm that our algorithm gives good estimates of the 3D hand pose. Keywords: 3D hand posture estimation; Model-based approach; Gesture recognition; human- computer interface; machine vision.
Camera as Mouse and Keyboard for Handicap Person with Troubleshooting Ability...Waqas Tariq
Camera mouse has been widely used for handicap person to interact with computer. The utmost important of the use of camera mouse is must be able to replace all roles of typical mouse and keyboard. It must be able to provide all mouse click events and keyboard functions (include all shortcut keys) when it is used by handicap person. Also, the use of camera mouse must allow users troubleshooting by themselves. Moreover, it must be able to eliminate neck fatigue effect when it is used during long period. In this paper, we propose camera mouse system with timer as left click event and blinking as right click event. Also, we modify original screen keyboard layout by add two additional buttons (button “drag/ drop” is used to do drag and drop of mouse events and another button is used to call task manager (for troubleshooting)) and change behavior of CTRL, ALT, SHIFT, and CAPS LOCK keys in order to provide shortcut keys of keyboard. Also, we develop recovery method which allows users go from camera and then come back again in order to eliminate neck fatigue effect. The experiments which involve several users have been done in our laboratory. The results show that the use of our camera mouse able to allow users do typing, left and right click events, drag and drop events, and troubleshooting without hand. By implement this system, handicap person can use computer more comfortable and reduce the dryness of eyes.
A Proposed Web Accessibility Framework for the Arab DisabledWaqas Tariq
The Web is providing unprecedented access to information and interaction for people with disabilities. This paper presents a Web accessibility framework which offers the ease of the Web accessing for the disabled Arab users and facilitates their lifelong learning as well. The proposed framework system provides the disabled Arab user with an easy means of access using their mother language so they don’t have to overcome the barrier of learning the target-spoken language. This framework is based on analyzing the web page meta-language, extracting its content and reformulating it in a suitable format for the disabled users. The basic objective of this framework is supporting the equal rights of the Arab disabled people for their access to the education and training with non disabled people. Key Words : Arabic Moon code, Arabic Sign Language, Deaf, Deaf-blind, E-learning Interactivity, Moon code, Web accessibility , Web framework , Web System, WWW.
Real Time Blinking Detection Based on Gabor FilterWaqas Tariq
New method of blinking detection is proposed. The utmost important of blinking detections method is robust against different users, noise, and also change of eye shape. In this paper, we propose blinking detections method by measuring of distance between two arcs of eye (upper part and lower part). We detect eye arcs by apply Gabor filter onto eye image. As we know that Gabor filter has advantage on image processing application since it able to extract spatial localized spectral features, such line, arch, and other shape are more easily detected. After two of eye arcs are detected, we measure the distance between both by using connected labeling method. The open eye is marked by the distance between two arcs is more than threshold and otherwise, the closed eye is marked by the distance less than threshold. The experiment result shows that our proposed method robust enough against different users, noise, and eye shape changes with perfectly accuracy.
Computer Input with Human Eyes-Only Using Two Purkinje Images Which Works in ...Waqas Tariq
A method for computer input with human eyes-only using two Purkinje images which works in a real time basis without calibration is proposed. Experimental results shows that cornea curvature can be estimated by using two light sources derived Purkinje images so that no calibration for reducing person-to-person difference of cornea curvature. It is found that the proposed system allows usersf movements of 30 degrees in roll direction and 15 degrees in pitch direction utilizing detected face attitude which is derived from the face plane consisting three feature points on the face, two eyes and nose or mouth. Also it is found that the proposed system does work in a real time basis.
Toward a More Robust Usability concept with Perceived Enjoyment in the contex...Waqas Tariq
Mobile multimedia service is relatively new but has quickly dominated people¡¯s lives, especially among young people. To explain this popularity, this study applies and modifies the Technology Acceptance Model (TAM) to propose a research model and conduct an empirical study. The goal of study is to examine the role of Perceived Enjoyment (PE) and what determinants can contribute to PE in the context of using mobile multimedia service. The result indicates that PE is influencing on Perceived Usefulness (PU) and Perceived Ease of Use (PEOU) and directly Behavior Intention (BI). Aesthetics and flow are key determinants to explain Perceived Enjoyment (PE) in mobile multimedia usage.
Collaborative Learning of Organisational KnolwedgeWaqas Tariq
This paper presents recent research into methods used in Australian Indigenous Knowledge sharing and looks at how these can support the creation of suitable collaborative envi- ronments for timely organisational learning. The protocols and practices as used today and in the past by Indigenous communities are presented and discussed in relation to their relevance to a personalised system of knowledge sharing in modern organisational cultures. This research focuses on user models, knowledge acquisition and integration of data for constructivist learning in a networked repository of or- ganisational knowledge. The data collected in the repository is searched to provide collections of up-to-date and relevant material for training in a work environment. The aim is to improve knowledge collection and sharing in a team envi- ronment. This knowledge can then be collated into a story or workflow that represents the present knowledge in the organisation.
Our research aims to propose a global approach for specification, design and verification of context awareness Human Computer Interface (HCI). This is a Model Based Design approach (MBD). This methodology describes the ubiquitous environment by ontologies. OWL is the standard used for this purpose. The specification and modeling of Human-Computer Interaction are based on Petri nets (PN). This raises the question of representation of Petri nets with XML. We use for this purpose, the standard of modeling PNML. In this paper, we propose an extension of this standard for specification, generation and verification of HCI. This extension is a methodological approach for the construction of PNML with Petri nets. The design principle uses the concept of composition of elementary structures of Petri nets as PNML Modular. The objective is to obtain a valid interface through verification of properties of elementary Petri nets represented with PNML.
Development of Sign Signal Translation System Based on Altera’s FPGA DE2 BoardWaqas Tariq
The main aim of this paper is to build a system that is capable of detecting and recognizing the hand gesture in an image captured by using a camera. The system is built based on Altera’s FPGA DE2 board, which contains a Nios II soft core processor. Image processing techniques and a simple but effective algorithm are implemented to achieve this purpose. Image processing techniques are used to smooth the image in order to ease the subsequent processes in translating the hand sign signal. The algorithm is built for translating the numerical hand sign signal and the result are displayed on the seven segment display. Altera’s Quartus II, SOPC Builder and Nios II EDS software are used to construct the system. By using SOPC Builder, the related components on the DE2 board can be interconnected easily and orderly compared to traditional method that requires lengthy source code and time consuming. Quartus II is used to compile and download the design to the DE2 board. Then, under Nios II EDS, C programming language is used to code the hand sign translation algorithm. Being able to recognize the hand sign signal from images can helps human in controlling a robot and other applications which require only a simple set of instructions provided a CMOS sensor is included in the system.
An overview on Advanced Research Works on Brain-Computer InterfaceWaqas Tariq
A brain–computer interface (BCI) is a proficient result in the research field of human- computer synergy, where direct articulation between brain and an external device occurs resulting in augmenting, assisting and repairing human cognitive. Advanced works like generating brain-computer interface switch technologies for intermittent (or asynchronous) control in natural environments or developing brain-computer interface by Fuzzy logic Systems or by implementing wavelet theory to drive its efficacies are still going on and some useful results has also been found out. The requirements to develop this brain machine interface is also growing day by day i.e. like neuropsychological rehabilitation, emotion control, etc. An overview on the control theory and some advanced works on the field of brain machine interface are shown in this paper.
Exploring the Relationship Between Mobile Phone and Senior Citizens: A Malays...Waqas Tariq
There is growing ageing phenomena with the rise of ageing population throughout the world. According to the World Health Organization (2002), the growing ageing population indicates 694 million, or 223% is expected for people aged 60 and over, since 1970 and 2025.The growth is especially significant in some advanced countries such as North America, Japan, Italy, Germany, United Kingdom and so forth. This growing older adult population has significantly impact the social-culture, lifestyle, healthcare system, economy, infrastructure and government policy of a nation. However, there are limited research studies on the perception and usage of a mobile phone and its service for senior citizens in a developing nation like Malaysia. This paper explores the relationship between mobile phones and senior citizens in Malaysia from the perspective of a developing country. We conducted an exploratory study using contextual interviews with 5 senior citizens of how they perceive their mobile phones. This paper reveals 4 interesting themes from this preliminary study, in addition to the findings of the desirable mobile requirements for local senior citizens with respect of health, safety and communication purposes. The findings of this study bring interesting insight to local telecommunication industries as a whole, and will also serve as groundwork for more in-depth study in the future.
Principles of Good Screen Design in WebsitesWaqas Tariq
Visual techniques for proper arrangement of the elements on the user screen have helped the designers to make the screen look good and attractive. Several visual techniques emphasize the arrangement and ordering of the screen elements based on particular criteria for best appearance of the screen. This paper investigates few significant visual techniques in various web user interfaces and showcases the results for better understanding and their presence.
Virtual teams are used more and more by companies and other organizations to receive benefits. They are a great way to enable teamwork in situations where people are not sitting in the same physical place at the same time. As companies seek to increase the use of virtual teams, a need exists to explore the context of these teams, the virtuality of a team and software that may help these teams working virtualy. Virtual teams have the same basic principles as traditional teams, but there is one big difference. This difference is the way the team members communicate. Instead of using the dynamics of in-office face-to-face exchange, they now rely on special communication channels enabled by modern technologies, such as e-mails, faxes, phone calls and teleconferences, virtual meetings etc. This is why this paper is focused on the issues regarding virtual teams, and how these teams are created and progressing in Albania.
Cognitive Approach Towards the Maintenance of Web-Sites Through Quality Evalu...Waqas Tariq
It is a well established fact that the Web-Applications require frequent maintenance because of cutting– edge business competitions. The authors have worked on quality evaluation of web-site of Indian ecommerce domain. As a result of that work they have made a quality-wise ranking of these sites. According to their work and also the survey done by various other groups Futurebazaar web-site is considered to be one of the best Indian e-shopping sites. In this research paper the authors are assessing the maintenance of the same site by incorporating the problems incurred during this evaluation. This exercise gives a real world maintainability problem of web-sites. This work will give a clear picture of all the quality metrics which are directly or indirectly related with the maintainability of the web-site.
USEFul: A Framework to Mainstream Web Site Usability through Automated Evalua...Waqas Tariq
A paradox has been observed whereby web site usability is proven to be an essential element in a web site, yet at the same time there exist an abundance of web pages with poor usability. This discrepancy is the result of limitations that are currently preventing web developers in the commercial sector from producing usable web sites. In this paper we propose a framework whose objective is to alleviate this problem by automating certain aspects of the usability evaluation process. Mainstreaming comes as a result of automation, therefore enabling a non-expert in the field of usability to conduct the evaluation. This results in reducing the costs associated with such evaluation. Additionally, the framework allows the flexibility of adding, modifying or deleting guidelines without altering the code that references them since the guidelines and the code are two separate components. A comparison of the evaluation results carried out using the framework against published evaluations of web sites carried out by web site usability professionals reveals that the framework is able to automatically identify the majority of usability violations. Due to the consistency with which it evaluates, it identified additional guideline-related violations that were not identified by the human evaluators.
Robot Arm Utilized Having Meal Support System Based on Computer Input by Huma...Waqas Tariq
A robot arm utilized having meal support system based on computer input by human eyes only is proposed. The proposed system is developed for handicap/disabled persons as well as elderly persons and tested with able persons with several shapes and size of eyes under a variety of illumination conditions. The test results with normal persons show the proposed system does work well for selection of the desired foods and for retrieve the foods as appropriate as usersf requirements. It is found that the proposed system is 21% much faster than the manually controlled robotics.
Dynamic Construction of Telugu Speech Corpus for Voice Enabled Text EditorWaqas Tariq
In recent decades speech interactive systems have gained increasing importance. Performance of an ASR system mainly depends on the availability of large corpus of speech. The conventional method of building a large vocabulary speech recognizer for any language uses a top-down approach to speech. This approach requires large speech corpus with sentence or phoneme level transcription of the speech utterances. The transcriptions must also include different speech order so that the recognizer can build models for all the sounds present. But, for Telugu language, because of its complex nature, a very large, well annotated speech database is very difficult to build. It is very difficult, if not impossible, to cover all the words of any Indian language, where each word may have thousands and millions of word forms. A significant part of grammar that is handled by syntax in English (and other similar languages) is handled within morphology in Telugu. Phrases including several words (that is, tokens) in English would be mapped on to a single word in Telugu.Telugu language is phonetic in nature in addition to rich in morphology. That is why the speech technology developed for English cannot be applied to Telugu language. This paper highlights the work carried out in an attempt to build a voice enabled text editor with capability of automatic term suggestion. Main claim of the paper is the recognition enhancement process developed by us for suitability of highly inflecting, rich morphological languages. This method results in increased speech recognition accuracy with very much reduction in corpus size. It also adapts Telugu words to the database dynamically, resulting in growth of the corpus.
An Improved Approach for Word Ambiguity RemovalWaqas Tariq
Word ambiguity removal is a task of removing ambiguity from a word, i.e. correct sense of word is identified from ambiguous sentences. This paper describes a model that uses Part of Speech tagger and three categories for word sense disambiguation (WSD). Human Computer Interaction is very needful to improve interactions between users and computers. For this, the Supervised and Unsupervised methods are combined. The WSD algorithm is used to find the efficient and accurate sense of a word based on domain information. The accuracy of this work is evaluated with the aim of finding best suitable domain of word. Keywords: Human Computer Interaction, Supervised Training, Unsupervised Learning, Word Ambiguity, Word sense disambiguation
Parameters Optimization for Improving ASR Performance in Adverse Real World N...Waqas Tariq
From the existing research it has been observed that many techniques and methodologies are available for performing every step of Automatic Speech Recognition (ASR) system, but the performance (Minimization of Word Error Recognition-WER and Maximization of Word Accuracy Rate- WAR) of the methodology is not dependent on the only technique applied in that method. The research work indicates that, performance mainly depends on the category of the noise, the level of the noise and the variable size of the window, frame, frame overlap etc is considered in the existing methods. The main aim of the work presented in this paper is to use variable size of parameters like window size, frame size and frame overlap percentage to observe the performance of algorithms for various categories of noise with different levels and also train the system for all size of parameters and category of real world noisy environment to improve the performance of the speech recognition system. This paper presents the results of Signal-to-Noise Ratio (SNR) and Accuracy test by applying variable size of parameters. It is observed that, it is really very hard to evaluate test results and decide parameter size for ASR performance improvement for its resultant optimization. Hence, this study further suggests the feasible and optimum parameter size using Fuzzy Inference System (FIS) for enhancing resultant accuracy in adverse real world noisy environmental conditions. This work will be helpful to give discriminative training of ubiquitous ASR system for better Human Computer Interaction (HCI). Keywords: ASR Performance, ASR Parameters Optimization, Multi-Environmental Training, Fuzzy Inference System for ASR, ubiquitous ASR system, Human Computer Interaction (HCI)
2024.06.01 Introducing a competency framework for languag learning materials ...Sandy Millin
http://sandymillin.wordpress.com/iateflwebinar2024
Published classroom materials form the basis of syllabuses, drive teacher professional development, and have a potentially huge influence on learners, teachers and education systems. All teachers also create their own materials, whether a few sentences on a blackboard, a highly-structured fully-realised online course, or anything in between. Despite this, the knowledge and skills needed to create effective language learning materials are rarely part of teacher training, and are mostly learnt by trial and error.
Knowledge and skills frameworks, generally called competency frameworks, for ELT teachers, trainers and managers have existed for a few years now. However, until I created one for my MA dissertation, there wasn’t one drawing together what we need to know and do to be able to effectively produce language learning materials.
This webinar will introduce you to my framework, highlighting the key competencies I identified from my research. It will also show how anybody involved in language teaching (any language, not just English!), teacher training, managing schools or developing language learning materials can benefit from using the framework.
Students, digital devices and success - Andreas Schleicher - 27 May 2024..pptxEduSkills OECD
Andreas Schleicher presents at the OECD webinar ‘Digital devices in schools: detrimental distraction or secret to success?’ on 27 May 2024. The presentation was based on findings from PISA 2022 results and the webinar helped launch the PISA in Focus ‘Managing screen time: How to protect and equip students against distraction’ https://www.oecd-ilibrary.org/education/managing-screen-time_7c225af4-en and the OECD Education Policy Perspective ‘Students, digital devices and success’ can be found here - https://oe.cd/il/5yV
The French Revolution, which began in 1789, was a period of radical social and political upheaval in France. It marked the decline of absolute monarchies, the rise of secular and democratic republics, and the eventual rise of Napoleon Bonaparte. This revolutionary period is crucial in understanding the transition from feudalism to modernity in Europe.
For more information, visit-www.vavaclasses.com
Instructions for Submissions thorugh G- Classroom.pptxJheel Barad
This presentation provides a briefing on how to upload submissions and documents in Google Classroom. It was prepared as part of an orientation for new Sainik School in-service teacher trainees. As a training officer, my goal is to ensure that you are comfortable and proficient with this essential tool for managing assignments and fostering student engagement.
The Art Pastor's Guide to Sabbath | Steve ThomasonSteve Thomason
What is the purpose of the Sabbath Law in the Torah. It is interesting to compare how the context of the law shifts from Exodus to Deuteronomy. Who gets to rest, and why?
Read| The latest issue of The Challenger is here! We are thrilled to announce that our school paper has qualified for the NATIONAL SCHOOLS PRESS CONFERENCE (NSPC) 2024. Thank you for your unwavering support and trust. Dive into the stories that made us stand out!
Operation “Blue Star” is the only event in the history of Independent India where the state went into war with its own people. Even after about 40 years it is not clear if it was culmination of states anger over people of the region, a political game of power or start of dictatorial chapter in the democratic setup.
The people of Punjab felt alienated from main stream due to denial of their just demands during a long democratic struggle since independence. As it happen all over the word, it led to militant struggle with great loss of lives of military, police and civilian personnel. Killing of Indira Gandhi and massacre of innocent Sikhs in Delhi and other India cities was also associated with this movement.
We all have good and bad thoughts from time to time and situation to situation. We are bombarded daily with spiraling thoughts(both negative and positive) creating all-consuming feel , making us difficult to manage with associated suffering. Good thoughts are like our Mob Signal (Positive thought) amidst noise(negative thought) in the atmosphere. Negative thoughts like noise outweigh positive thoughts. These thoughts often create unwanted confusion, trouble, stress and frustration in our mind as well as chaos in our physical world. Negative thoughts are also known as “distorted thinking”.
Affect Sensing and Contextual Affect Modeling from Improvisational Interaction
1. Li Zhang
International Journal of Computational Linguistics (IJCL), Volume (1): Issue (4) 45
Affect Sensing and Contextual Affect Modeling from
Improvisational Interaction
Li Zhang l.zhang@tees.ac.uk
School of Computing
Teesside University
Middlesbrough, TS1 3BA, UK
Abstract
We report work on adding an improvisational AI actor to an existing virtual
improvisational environment, a text-based software system for dramatic
improvisation in simple virtual scenarios, for use primarily in learning contexts.
The improvisational AI actor has an affect-detection component, which is aimed
at detecting affective aspects (concerning emotions, moods, value judgments,
etc.) of human-controlled characters’ textual “speeches”. The AI actor will also
make an appropriate response based on this affective understanding, which
intends to stimulate the improvisation. The work also accompanies basic
research into how affect is conveyed linguistically. A distinctive feature of the
project is a focus on the metaphorical ways in which affect is conveyed.
Moreover, we have also introduced affect detection using context profiles.
Finally, we have reported user testing conducted for the improvisational AI actor
and evaluation results of the affect detection component. Our work contributes to
the journal themes on affective user interfaces, affect sensing and
improvisational or dramatic natural language interaction.
Keywords: Affect detection, metaphorical language, intelligent conversational agents, dramatic
improvisation and context profiles.
1. INTRODUCTION
In our previous work, we have developed online multi-user role-play software that could be used
for education or entertainment. In this software young people could interact online in a 3D virtual
drama stage with others under the guidance of a human director. In one session, up to five virtual
characters are controlled on a virtual stage by human users (“actors”), with characters’ (textual)
“speeches” typed by the actors operating the characters. A graphical interface on each actor’s
and director’s terminal shows the stage and characters. Speeches are shown as text bubbles.
Actors and the human director work through software clients connecting with the server. The
actors are given a loose scenario around which to improvise, but are at liberty to be creative.
The human director needs to constantly monitor the unfolding drama and the actors’ interactions,
or lack of them, in order to check whether they are keeping to the general spirit of the scenario. If
this is not happening, the director may then intervene. Director’s interventions may take a number
of forms. Director may choose to send messages to actors or may introduce and control a bit-part
character. This character may not have a major role in the drama, but can help to stimulate the
improvisation. But this places a heavy burden on directors, especially if they are, for example,
teachers and unpracticed in the directorial role.
2. Li Zhang
International Journal of Computational Linguistics (IJCL), Volume (1): Issue (4) 46
One research aim is thus partially to automate the directorial functions, which importantly involve
affect detection. For instance, a director may intervene when emotions expressed or discussed
by characters are not as expected. Hence we have developed an affect-detection module. The
module identifies affect in characters’ text input, and makes appropriate responses to help
stimulate the improvisation. Within affect we include: basic and complex emotions such as anger
and embarrassment; meta-emotions such as desiring to overcome anxiety; moods such as
hostility; and value judgments (of goodness, etc.). Although merely detecting affect is limited
compared to extracting full meaning, this is often enough for stimulating improvisation. The
results of this affective analysis are then used to: (a) control an automated improvisational AI
actor – EMMA (emotion, metaphor and affect) that operates a bit-part character (a minor
character) in the improvisation; (b) drive the animations of the avatars in the user interface so that
they react bodily in ways that is consistent with the affect that they are expressing, for instance by
changing posture or facial expressions.
Much research has been done on creating affective virtual characters in interactive systems.
Indeed, Picard’s work [1] makes great contributions to building affective virtual characters. Also,
emotion theories, particularly that of Ortony, et al. [2] (OCC), have been used widely in such
research. Egges et al. [3] provided virtual characters with conversational emotional
responsiveness. Aylett et al. [4] also focused on the development of affective behaviour planning
for their synthetic characters. There is much other work in a similar vein.
Emotion recognition in speech and facial expression has been studied extensively [5, 6]. But very
little research work has made an attempt to dig out the affect flavour in human open-ended
linguistic textual input in online role-play, although the first interaction system based on natural
language textual input, Eliza, was first developed back in 1966. Thus there has been only a
limited amount of work directly comparable to our own, especially given our concentration on
improvisation and open-ended language. However, Façade [7] included shallow natural language
processing for characters’ open-ended utterances. But the detection of major emotions, rudeness
and value judgements is not mentioned. Zhe and Boucouvalas [8] demonstrated an emotion
extraction module embedded in an Internet chatting environment. Unfortunately the emotion
detection focuses only on emotional adjectives, and does not address deep issues such as
figurative expression of emotion. Also, the concentration purely on first-person emotions is
narrow. Our work is thus distinctive in these aspects, including affect detection in metaphorical
language and context profiles, and also from first-person and third-person perspectives.
Various characterizations of emotion are used in emotion theories. The OCC model uses emotion
labels (anger, etc.) and intensity, while Watson and Tellegen [9] use positivity and negativity of
affect as the major dimensions. We have drawn ideas from several such sources. We use an
evaluation dimension (negative-positive), affect labels, and intensity. The basic emotion labels
(such as ‘angry’) we use are taken from Ekman [10], while other comparatively complex affect
labels (such as ‘approving’) are taken from the OCC model. There are 25 affect labels used in our
system currently. Affect labels plus intensity are used when strong text clues signalling affect are
detected, while the evaluation dimension plus intensity is used when only weak text clues are
detected. In this paper, although first we briefly summarize our previous implementation in section
2.1 & 2.2, we mainly emphasis our new implementations on metaphorical figurative language
processing in section 2.3, and affect interpretation based on context in section 2.4 and user
testing evaluation for the AI agent and the overall affect sensing component in section 3. We
draw conclusion and identify future work in section 4.
2. THE AFFECT DETECTION PROCESSING
Before any automatic recognition and response components could be built for use in our
automated AI actor, a detailed analysis of the language used in e-drama sessions was
necessary. A small corpus of sessions was analysed by hand to identify the range of linguistic
forms used and to provide insight for the automatic processing. In fact, this analysis is often very
difficult and unreliable but it does reveal some important observations. The language used in e-
drama is complex and idiosyncratic, e.g. often ungrammatical and full of abbreviations, mis-
3. Li Zhang
International Journal of Computational Linguistics (IJCL), Volume (1): Issue (4) 47
spellings, etc. Moreover, the language contains a large number of weak cues to the affect that is
being expressed. These cues may be contradictory or they may work together to enable a
stronger interpretation of the affective state. In order to build a reliable and robust analyser of
affect it is necessary to undertake several diverse forms of analysis and to enable these to work
together to build stronger interpretations.
2.1 Pre-processing Modules
The language created in e-drama sessions severely challenges existing language-analysis tools if
accurate semantic information is sought, even in the limited domain of restricted affect-detection.
Aside from the complications noted above, the language includes slang, use of upper case and
special punctuation (such as repeated exclamation marks) for affective emphasis, repetition of
letters, syllables or words for emphasis, and open-ended interjective and onomatopoeic elements
such as “hm”, “ow” and “grrrr”. To deal with the misspellings, abbreviations, letter repetitions,
interjections and onomatopoeia, several types of pre-processing occur before the main aspects of
detection of affect. We have reported our work on pre-processing modules to deal with these
language phenomena in detail in [11, 25].
2.2 Affect Detection using Rasp, Pattern Matching & WordNet and Responding Regimes
One useful pointer to affect is the use of imperative mood, especially when used without
softeners such as ‘please’ or ‘would you’. Strong emotions and/or rude attitudes are often
expressed in this case. Expression of the imperative mood in English is surprisingly various and
ambiguity-prone. We have used the syntactic output from the Rasp parser [12] and semantic
information in the form of the semantic profiles for the 1,000 most frequently used English words
[13] to deal with certain types of imperatives.
In an initial stage of our work, affect detection was based purely on textual pattern-matching rules
that looked for simple grammatical patterns or templates partially involving specific words or sets
of specific alternative words. This continues to be a core aspect of our system but we have now
added robust parsing and some semantic analysis, including but going beyond the handling of
imperatives discussed above.
A rule-based Java framework called Jess is used to implement the pattern/template-matching
rules in the AI agent allowing the system to cope with more general wording. This procedure
possesses the robustness and flexibility to accept many ungrammatical fragmented sentences.
The rules conjecture the character’s emotions, evaluation dimension (negative or positive),
politeness (rude or polite) and what response the automated actor should make. However, it
lacks other types of generality and can be fooled when the phrases are suitably embedded as
subcomponents of other grammatical structures. In order to go beyond certain such limitations,
sentence type information obtained from the Rasp parser has also been adopted in the pattern-
matching rules. This information not only helps the AI agent to detect affective states in the user’s
input (such as the detection of imperatives), and to decide if the detected affective states should
be counted (e.g. affects detected in conditional sentences won’t be valued), but also contributes
to proposing appropriate responses.
Additionally, a reasonably good indicator that an inner state is being described is the use of ‘I’,
especially in combination with the present or future tense (e.g. ‘I’ll scream’, ‘I hate/like you’, and ‘I
need your help’). We especially process ‘the first-person with a present-tense verb’ statements
using WordNet.
We have also created responding regimes for the AI character. Most importantly, the AI agent
can adjust its response likelihood according to how confident the AI agent is about what it has
discerned in the utterance at hand. Especially, in order to make contributions to the improvisation
progression, the AI agent also has a global view of the drama improvisation. Briefly, the
knowledge base of the AI actor provides scenario’s background knowledge for each human
character. The AI agent can raise various scenario-related topics in its role for the human
characters according to the detected affective states and topics discussed in the text input by
4. Li Zhang
International Journal of Computational Linguistics (IJCL), Volume (1): Issue (4) 48
using the rule-based reasoning based on the knowledge base. Inspection of the transcripts
collected in the user testing indicates that the AI actor usefully pushed the improvisation forward
on various occasions (see section 3). Details of the work reported in this section can be found in
[11, 25].
2.3 Metaphorical Language Understanding in the AI Actor
The metaphorical description of emotional states is common and has been extensively studied
[14, 15]. E.g.: “He nearly exploded” and “Joy ran through me,” where anger and joy are being
viewed in vivid physical terms. Such examples describe emotional states in a relatively explicit if
metaphorical way. But affect is also often conveyed more implicitly via metaphor, as in “His room
is a cess-pit”: affect (such as ‘disgust’) associated with a source item (cess-pit) gets carried over
to the corresponding target item (the room). In other work, we have conducted research on
metaphor in general (see, e.g. [16, 17]), and are now applying it to this application, and
conversely using the application as a useful source of theoretical inspiration.
In our collected transcripts, metaphorical language has been used extensively to convey
emotions and feelings. One category of affective metaphorical expressions that we’re interested
in is ‘Ideas/Emotions as Physical Objects” [16, 17], e.g. “joy ran through me”, “my anger returns in
a rush”, “fear is killing me” etc. In these examples, emotions and feelings have been regarded as
external entities. The external entities are often, or usually, physical objects or events. Therefore,
affects could be treated as physical objects outside the agent in such examples, which could be
active in other ways [16]. Implementation has been carried out to provide the affect detection
component the ability to deal with such affect metaphor. We mainly focus on the user input with
the following structures: ‘a singular common noun subject + present-tense lexical verb phrase’ or
‘a singular common noun subject + present-tense copular form + -ing form of lexical verb phrase’.
WordNet-affect domain (part of WordNet-domain 3.2) [18] has been used in our application. It
provides an additional hierarchy of ‘affective domain labels’, with which the synsets representing
affective concepts are further annotated (e.g. ‘panic’ is interpreted as ‘negative-fear -> negative-
emotion -> emotion -> affective-state -> mental-state’). Also with the assistance of the syntactic
parsing from Rasp, the input “panic drags me down” is interpreted as ‘a mental state + an activity
+ object (me)’. Thus the system regards such expression as affective metaphor belonging to the
category of ‘affects as entities’.
In daily expressions, food has been used extensively as metaphor for social position, group
identity, religion, etc. E.g. food could also be used as a metaphor for national identity. British have
been called ‘roastbeefs’ by the French, while French have been referred to as ‘frogs’ by the
British. In one of the scenarios we used (school bullying), the big bully has called the bullied
victim (Lisa) names, such as “u r a pizza”, “Lisa has a pizza face” to exaggerate that fact that the
victim has acne. Another most commonly used food metaphor is to use food to refer to a specific
shape. E.g. body shape could be described as ‘banana’, ‘pear’ and ‘apple’
(http://jsgfood.blogspot.com/2008/02/food-metaphors.html). In our application, “Lisa has a pizza
face” could also be interpreted as Lisa has a ‘round (shape)’ face. Therefore, insults could be
conveyed in such food metaphorical expression. We especially focus on the statement of
‘second-person/a singular proper noun + present-tense copular form + food term’ to extract affect.
A special semantic dictionary has been created by providing semantic tags to normal English
lexicon. The semantic tags have been created by using Wmatrix [19], which facilitates the user to
obtain corpus annotation with semantic and part-of-speech tags to compose dictionary. The
semantic dictionary created consists mainly of food terms, animal names, measureable adjectives
(such as size) etc with their corresponding semantic tags due to the fact they have the potential to
convey affect and feelings.
In our application, Rasp informs the system the user input with the desired structure - ‘second-
person/a singular proper noun + present-tense copular form + noun phrases’ (e.g. “Lisa is a
pizza”, “u r a hard working man”, “u r a peach”). The noun phrases are examined in order to
recover the main noun term. Then its corresponding semantic tag is derived from the composed
semantic dictionary if it is a food term, or an animal-name etc. E.g. “u r a peach” has been
5. Li Zhang
International Journal of Computational Linguistics (IJCL), Volume (1): Issue (4) 49
regarded as “second-person + present-tense copular form + [food-term]”. WordNet [20] has been
employed in order to get the synset of the food term. If among the synset, the food term has been
explained as a certain type of human being, such as ‘beauty’, ‘sweetheart’ etc. Then another
small slang-semantic dictionary collected in our previous study containing terms for special
person types (such as ‘freak’, ‘angle’) and their corresponding evaluation values (negative or
positive) has been adopted in order to obtain the evaluation values of such synonyms of the food
term. If the synonyms are positive (e.g. ‘beauty’), then we conclude that the input is an
affectionate expression with a food metaphor (e.g. “u r a peach”).
However, in most of the cases, WordNet doesn’t provide any description of types of human
beings when explaining a food term (e.g. ‘pizza’, ‘meat’ etc). According to the nature of the
scenarios (e.g. bullying) we used, we simply conclude that the input implies insulting with a food
metaphor when calling someone food terms (“u r walking meat”, “Lisa is a pizza”).
Another interesting phenomenon drawing our attention is food as shape metaphor. As mentioned
earlier, food is often used as a metaphor to refer to body shapes (e.g. “you have a pear body
shape”, “Lisa has a garlic nose”, “Lisa has a pizza face”). They might indicate literal truth, but
most of which are potentially used to indicate very unpleasant truth. Thus they could be regarded
as insulting. We extend our semantic dictionary created with the assistance of Wmatrix by adding
terms of physiological human body parts, such as face, nose, body etc. For the user’s input with a
structure of ‘second-person/a singular proper noun + have/has + noun phrases’ informed by
Rasp, the system provides a semantic tag for each word in the object noun phrase. If the
semantic tag sequence of the noun phrase indicates that it consists of a food term followed by a
physiological term (‘pizza face’), the system interprets that the input implies insulting with a food
metaphor.
However, examples, such as “you have a banana body shape” and “you are a meat and potatoes
man”, haven’t been used to express insults, but instead the former used to indicate a slim body
and the latter to indicate a hearty appetite and robust character. Other examples such as “you are
what you eat” could be very challenging theoretically and practically. In order to gain more
flexibility and generalization when dealing with metaphorical expressions, we have also used a
statistical-based machine learning approach to conduct some experiments on the recognition of
the above affect and food metaphors.
2.4 Context-based Affect Detection
Our previous affect detection has been performed solely based on individual turn-taking user
input. Thus the context information has been ignored. However, the contextual and character
profiles may influence the affect conveyed in the current input. In this section, we are going to
discuss cognitive emotion simulation for individual characters and contextual emotion modeling
for other characters’ influence towards the current speaking character in communication context
and our approach developed based on these features to interpret affect from context.
In our study, we previously noticed some linguistic indicators for contextual communication in the
recorded transcripts. E.g. one useful indicator is (i) imperatives, which are often used to imply
negative or positive responses to the previous speaking characters, such as “shut up”, “go on
then”, “let’s do it” and “bring it on”. Other useful contextual indicators are (ii) prepositional phrases
(e.g. “by who?”), semi-coordinating conjunctions (e.g. “so we are good then”), subordinating
conjunctions (“because Lisa is a dog”) and coordinating conjunctions (‘and’, ‘or’ and ‘but’). These
indicators are normally used by the current ‘speaker’ to express further opinions or gain further
confirmation from the previous speakers.
In addition, (iii) short phrases for questions are also used frequently in the transcripts to gain
further communication based on context, e.g. “where?”, “who is Dave” or “what”. (iv) Character
names are also normally used in the user input to indicate that the current input is intended for
particular characters, e.g. “Dave go away”, “Mrs Parton, say something”, “Dave what has got into
you?” etc. Very often, such expressions have been used to imply potential emotional contextual
6. Li Zhang
International Journal of Computational Linguistics (IJCL), Volume (1): Issue (4) 50
communication between the current speaking character and the named character. Therefore the
current speaking characters may imply at least ‘approval’ or ‘disapproval’ towards the
opinions/comments provided by the previous named speaking characters. Finally there are also
(i) some other well known contextual indicators in Internet relay chat such as ‘yeah/yes followed
by a sentence (“yeah, we will see”, “yeah, we cool Lisa”)’, “I think so”, ‘no/nah followed by a
sentence’, “me too”, “exactly”, “thanks”, “sorry”, “grrrr”, “hahahaha”, etc. Such expressions are
normally used to indicate affective responses to the previous input. However, these linguistic
indicators act as very limited signals for contextual communication. There are still cases (“ur a
batty 2 then okay”, “the rest dropped out cuz they didn’t want to play with a gay”, “I want to talk
about it now”) that contextual affect analysis fails to be activated to derive affect implied in the
user’s input. In the work reported here, we intend to deal with such difficulties by activating
contextual affect analysis even for input with structures of “subjects + verb phrases + objects”.
Especially an input with a structure of ‘second person + copular form (you are)’ tends to convey
insulting in our application (“u r a batty 2 then okay”, “u r walking meat” etc).
2.4.1 Emotion Modeling using Bayesian Networks
Lopez et al. [26] has suggested in their work that context profiles for affect detection have been
referred to social, environmental and personal contexts. In our study, personal context may be
regarded as one’s own emotion inclination or improvisational mood in communication context.
Bayesian networks have been used to simulate such personal emotion context. E.g. in this
Bayesian network, we regard the first emotion experienced by a user as A, the second emotion
experienced by the same user as B, and the third emotion experienced as C. We believe that
one’s own emotional states have a chain reaction effect. For example, the previous emotional
status may influence later emotional experience. We have made attempts to embody such chain
effects into emotion modeling for personal context. We assume that the second emotional state
B, in any combination is dependent on the first emotional state A. Further, we assume that the
third emotional state C, is dependent on both the first and second emotional states A and B. In
our application, if we only consider two most recent emotional states the user experiences as the
most related relevant context based on Relevance theory [21, 22], then we may predict what the
most probable emotion the user is the most likely to experience in the next turn-taking using a
Bayesian network.
A Bayesian network employs a probabilistic graphical model to represent causality relationship
and conditional (in)dependencies between domain variables. It allows combining prior knowledge
about (in)dependencies among variables with observed training data via a directed acyclic graph.
It has a set of directed arcs linking pairs of nodes: an arc from a node X to a node Y means that X
(parent emotion) has a direct influence on Y (successive emotion). Such causal modeling
between variables reflects the chain effect of emotional experience. It uses the conditional
probability (P[B|A], P[C|A,B]) to reflect such influence between prior emotional experiences to
successive emotional expression. The following network topology has been used to model
personal contextual emotional profiles in our application.
FIGURE 1: An Emotion Network
In Figure 1, conditional probabilities are needed to be calculated for the emotional state C given
any combination of the emotional states A and B. Theoretically, emotional states A and B could
be any combination of potential emotional states. Similarly, since there could be several
A C
B
7. Li Zhang
International Journal of Computational Linguistics (IJCL), Volume (1): Issue (4) 51
emotional states considered as successive emotional state C, we have considered a conditional
probability for each potential successive emotional state. In our application, we have mainly
considered the following 10 most frequently used emotional states for the simulation of the
improvisational mood for a particular character in the Bayesian network: ‘neural’, ‘happy’,
‘approval’, ‘grateful’, ‘caring’, ‘disapproval’, ‘sad’, ‘scared’, ‘insulting’, and ‘angry’. Any combination
of the above emotional states could be used as prior emotional experience of the user. Altogether
the overall combinations for the two prior emotions are counted as 100 (10 * 10). Also each
conditional probability of each emotional state in the above given two prior emotional experiences
(such as P[happy| A,B], P[approval| A,B] etc) will be calculated as the confidence for later
selection. Then the emotional state with the highest conditional probability, P[C|A,B], will be
chosen as the most probable emotional experience the user may express in his/her very next
turn-taking. In this way, we model contextual emotional chain effect for an individual character to
benefit our contextual affect detection.
An advantage of using Bayesian networks for emotion simulation and modeling is that it is not
necessary to gather training data from other sessions of the same scenarios to train the system at
the beginning to allow future prediction. We can simply use the emotional states experienced by a
particular character throughout the improvisation as the prior input emotions to the Bayesian
network so that our system may learn about this user’s emotional trend and mood gradually
without any constrains set by the training data or scenario related information.
Moreover we also take a frequency approach to determine the conditional probabilities. When an
affect has been detected from the user’s input, we increment a counter for that expressed
emotion given the two prior implied emotional states. An example conditional probability table has
been shown in Table 1.
Probability of the predicted emotional state
C being:
Emotion A Emotion B Happy Approval ... Angry
Happy Neutral P00 P01 ... P09
Neutral Angry P10 P11 ... P19
Disapproval Disapproval P20 P21 ... P29
Angry Angry P30 P31 ... P39
TABLE 1: Conditional Probability Table for Emotions Expressed
When making a prediction for an emotion state mostly likely to be shown in the very next input for
one particular character, the two prior emotional states are used to determine which row to
consider in the conditional probability matrix, and select the column with the highest conditional
probability as the final output. Example conditional probability calculations are shown in the
following formulas, where N represents the total number of emotions shown so far by this
particular character and N with a subscript indicates the number of a particular emotion shown
given previously expressed emotions. E.g., Nhappy_neutral_happy indicates the occurrences that two
prior emotions A and B are respectively happy & neutral and the following emotional state C is
happy.
P(A = happy) = Nhappy/N
P(B = neutral) = Nneutral/N
P(B = neutral| A = happy) = Nneutal_happy/N
P(C = happy| A = happy, B = neutral) = Nhappy_neutral_happy/NAB
As we mentioned earlier, the probabilities are not necessarily to be produced by any training data
and stored in advance. The frequencies are sufficient to use to calculate probabilities when
required. In our case, we store the frequencies of emotion combinations in a 100 * 10 ((A*B) * C)
matrix dynamically.
8. Li Zhang
International Journal of Computational Linguistics (IJCL), Volume (1): Issue (4) 52
In our application, one of the scenarios has been used for user testing is Homophobic bullying.
We briefly introduce this scenario in the following since example transcripts have been taken from
this scenario for the discussion of the contextual affect detection implementation reported here.
The character Dean (16 years old), captain of the football team, is confused about his sexuality.
He has ended a relationship with a girlfriend because he thinks he may be gay and has told her
this in confidence. Tiffany (ex-girlfriend) has told the whole school and now Dean is being bullied
and concerned that his team mates on the football team will react badly. He thinks he may have
to leave the team. The other characters are; Tiffany who is the ring leader of the bullying, and
wants Dean to leave the football team, Rob (Dean’s younger brother) who wants Dean to say he
is not gay to stop the bullying, Lea (Dean’s older sister) who wants Dean to be proud of who he is
and ignore the bullying, and Mr Dhanda (PE Teacher) who needs to confront Tiffany and stop the
bullying.
Suppose we have the following sequence of example interaction extracted from the recorded
transcripts for the Tiffany character in this scenario. Based on the affect detection purely from the
analysis of each individual input, we obtained the emotional states implied in the first three inputs
from Tiffany as the following: ‘angry, angry, and angry’.
Tiffany: Dean, U R DISGUSTING!! u shuld leav da football team. [angry]
...
Tiffany: shut up man lea [angry]
...
Tiffany: u get out of here. [angry]
...
Tiffany: ur a batty 2 then okay [neutral] -> [angry]
Also we have derived ‘neutral’ for the very last input without any contextual inference. Since the
input has a structure of “second person + copular form”, as discussed earlier which is very often
used to convey insulting or compliment in our application, the context-based affect analysis will
be activated to adjust/predict the affect conveyed in the last input from the above example
transcript. This emotional sequence implied by Tiffany (‘angry, angry, and angry’) will be used to
‘train’ the contextual emotional simulation and construct the Bayesian probability matrix, which
will be used to predict the most probable emotion implied in Tiffany’s very last input. In this
example, we need to calculate the conditional probability of P[C| angry, angry, angry] for each
potential emotional state C. Finally the emotional state ‘angry’ has achieved the highest
probability result and been predicted as the most probable emotion implied in the input “ur a batty
2 then okay”. Thus we adjust the emotional state for the very last input from ‘neutral’ to ‘angry’.
Therefore in this way, we can produce emotion modeling for each individual character within the
same and across scenarios. However, other contextual profiles (such as other characters’
emotional profiles and discussion topics) are still needed to further justify the affect detected
using the above discussed Bayesian network approach. In the following section, we introduce
how social emotional contextual profiles are used to model emotional influence from other
characters to the current speaking character during the improvisation.
2.4.2 Emotional Social Context Modeling using Unsupervised Neural Networks
The simulation of one’s own emotional context and improvisational mood is important, but the
modeling of other characters’ emotional influence to the current speaking character is also crucial
for the accurate interpretation of the affect implied in the current input. For example, the
emotional context contributed by other participants, e.g. friend or enemy characters, may
(dramatically) affect the speaking character’s emotional expression in the next turn-taking in our
application. Moreover, a discussion topic or an improvisation is composed of the mixture of such
emotional sub-contexts. They take the overall forms of being positive or negative and have been
acted as the most relevant emotional social context to the current speaking character. If such
social positive/negative most relevant context could be recognized during the improvisation, it is
very helpful to justify the affective states detected from personal context modeling using the
9. Li Zhang
International Journal of Computational Linguistics (IJCL), Volume (1): Issue (4) 53
above discussed Bayesian approach. In order to recognize the positive/negative trend in the most
related sub-context contributed by (part of) participants, an unsupervised neural network learning
algorithm has been employed. I.e. Adaptive Resonance Theory-1 (ART1) has been used in our
application to derive general emotional implication (positive/negative/neutral) for the most recent
interaction context.
Generally, ART is a collection of models for unsupervised learning, which deals with object
identification and recognition generally as a result of the interaction of ‘top-down’ observer
expectations with ‘bottom-up’ sensory information. The ‘top-down’ template or prototype will be
used to compare with the actual features of an object as detected by the senses to produce
categorizations for the observed objects. ART-1 in particular has been used to resolve stability
and plasticity dilemma, i.e. the ability to maintain previously learned knowledge (‘stability’) while
still being capable of learning new information (‘plasticity’). Although it mainly accepts binary input
vectors, this is sufficient enough in our application currently. In our application, it would be
beneficial that the positive/negative context prediction modeling is capable of both retaining
previously learned information (e.g. the sensing of positive or negative context in a particular
scenario) and in the meantime integrating newly discovered knowledge (e.g. the sensing of such
context across different scenarios). Such capability may allow the emotional social context
modeling to perform across scenarios. Also, the ART-1 algorithm has an advanced ability to
create a new cluster when required with the assistance of a vigilance parameter. It may help to
determine when to cluster a feature vector to a ‘close’ cluster or when a new cluster is needed to
accommodate this feature vector.
In our application, we use the evaluation values (positive and negative) and neutralization of the
most recent several turn-taking as the input to ART-1. In detail, for each user input, we convert its
emotional implication into pure positive or negative and use three binary values (0 or 1) to
represent the three emotional implications: neutral, positive and negative. For example, for the
input from Arnold in the Crohn’s disease scenario (another scenario used in our application),
“dont boss me about wife [angry]” when the wife character, Janet, was too pushy towards the
husband character, Arnold. We have used ‘0 (neutral), 0 (positive), and 1 (negative)’ to indicate
the emotional inclination (‘angry’ -> ‘negative’) in the user input. Another example transcript taken
from the Homophobic bullying scenario is listed in the following.
1. Tiffany Tanktop: sorry, all io could hear was I'M A BIG GAY [insulting/angry]
2. Mr. Dhanda: TIFFANY I WILL....GET YOU EXPENDED IF YOU DONT FOLLOW MY
ORDERS! YOU HOMO-FOBIC [angry]
3. Rob Hfuhruhurr: tiffany is wierd lol y she spreadn rumors etc???? [disapproval]
4. Tiffany Tanktop: there not rumours...its the truth [disapproval]
5. Tiffany Tanktop: GGGGAAAYYYYY! [insulting/angry]
6. Mr. Dhanda: TIFFANY STOP IT NOW!!! [angry]
7. Mr. Dhanda: ILL BANG YOU [angry]
8. Rob Hfuhruhurr: god leav hm alone!!! [angry]
9. Tiffany Tanktop: ONCE A BATTY ALWAYS A BATTY [neutral] -> [angry]
For the very last input from Tiffany, we can only interpret ‘neutral’ based on the analysis of the
input itself without using any contextual inference although emotional states have been derived
for all the other input based on the analysis the input themselves. In order to further derive/justify
the affect conveyed in the very last ‘neutral’ input although there is no any linguistic indicator for
contextual communication existing, we resort to the prediction of the general emotional trend
using the most related interaction context contributed by several participant characters. Since
normally in one session, up to 5 characters are involved in the improvisation as mentioned
previously, except for the last input, we have taken the previous last four inputs to the current last
input as the most related context for prediction of the positive/negative inclination in the social
context. Thus we have taken the input from Rob (8th input), Mr Dhanda (7th and 6th input), and
Tiffany (5th input) for consideration and prediction. Since Tiffany implies ‘angry’ (binary value
combination for neutral, positive and negative: 001) by saying “GGGGAAAYYYYY!”, Mr Dhanda
10. Li Zhang
International Journal of Computational Linguistics (IJCL), Volume (1): Issue (4) 54
also indicating ‘angry’ (001) in both of his input: “TIFFANY STOP IT NOW!!!” and “ILL BANG
YOU”, followed by another ‘angry’ (001) input from Rob “god leav hm alone!!!”, we have used the
following feature vector to represent this most related discussion context: ‘001 001 001 001
(angry, angry, angry and angry)’. This feature vector is used as the input to the ART-1
unsupervised learning algorithm to determine its category belongingness. In a similar way, we
can gather a set of such feature vectors from the same and across scenarios. ART-1 will classify
these feature vectors into different groups based on their similarities and differences shown in the
vectors, which sometimes may not be apparent at the beginning.
Briefly, we begin the ART-1 algorithm with a set of unclustered emotional context feature vectors
and some number of clusters. For each emotional feature vector, ART-1 makes attempts to find
the cluster to which it’s closest. A similarity test calculates how close the feature vector to the
cluster vector. The higher the value, the closer the vector is to the cluster. If a feature vector is
sufficiently close to a cluster, we then test for vigilance acceptability, which is the final determiner
for whether the feature vector should be added to the particular cluster. If a feature vector also
passes the vigilance test, then we assign it to and update that particular cluster with the features
of the new addition. If a feature vector fails the similarity test or vigilance test for all the available
clusters, then a new cluster is created for this feature vector. When new clusters are created,
some feature vectors may drop out of a cluster and into another based on new feature vectors
being added and adjusting the cluster vector. Thus ART-1 will start the process again by checking
through all the available feature vectors. If no feature vector needs to change its cluster, the
process is complete. In our application, we can gradually feed emotional context feature vectors
to ART-1, which will not only remain the previous classification of positive or negative context in a
particular scenario, but also indefinitely integrate new positive/negative context extracted from
other interaction across scenarios. Suppose we have the following emotional contexts contributed
by the Crohn’s disease scenario and classified previously by ART-1 into three categories:
Class 0 contains:
0 [1 0 0 0 0 1 0 0 1 0 0 1 ] negative1 (Neutral, sad, disapproving and sad)
1 [1 0 0 0 1 0 0 0 1 0 0 1 ] negative2 (Neutral, approving, disapproving and angry)
2 [1 0 0 0 0 1 0 0 1 1 0 0 ] negative3 (neutral, disapproving, disapproving and neutral)
3 [0 0 1 0 1 0 0 0 1 0 0 1 ] negative4 (angry, approving, disapproving, and angry)
5 [1 0 0 0 0 1 0 0 1 0 1 0 ] negative6 (neutral, angry, angry and approving)
Class 1 contains:
4 [0 0 1 0 0 1 1 0 0 1 0 0 ] negative5 (angry, angry, neutral and neutral)
8 [1 0 0 0 1 0 1 0 0 0 0 1 ] positive3 (neutral, caring, neutral and disapproval)
9 [1 0 0 1 0 0 1 0 0 1 0 0 ] neutral1 (neutral, neutral, neutral and neutral)
Class 2 contains:
6 [0 1 0 0 1 0 0 1 0 1 0 0 ] positive1 (happy, happy, happy and neutral)
7 [1 0 0 0 1 0 0 1 0 0 1 0 ] positive2 (neutral, caring, approving and happy)
10 [0 1 0 0 1 0 1 0 0 0 1 0] positive4 (approval, grateful, neutral and approval)
Since ART-1 is not aware which label it should use to mark the above each category although it
classifies the emotional feature vectors based on their similarities and differences and achieves
the above classification, a simple algorithm will make attempts to assign labels
(positive/negative/neutral context) to the above classification based on the majority vote of the
evaluation values of all the emotional states shown in each feature vector in each category. For
example, Class 0 has assigned 4 emotional feature vectors and most of the emotional states in
all the feature vectors in this category are ‘negative’, therefore it is labelled as ‘negative context’.
Similarly Class 1 is recognised as ‘neutral context’ with Class 2 identified as ‘positive context’. If
we add the above example emotional context from the Homophobic bullying scenario as a new
11. Li Zhang
International Journal of Computational Linguistics (IJCL), Volume (1): Issue (4) 55
feature vector, ‘001 001 001 001’ (angry, angry, angry and angry), to the algorithm, we have
Class 0 updated to accommodate the newly arrived feature vector as output. Thus the new
feature vector is ‘classified’ as ‘negative context’. Therefore, the last input from Tiffany (“ONCE A
BATTY ALWAYS A BATTY”) is more likely to contain ‘negative’ implication rather than ‘neutral’
based on the consideration of its most relevant emotional context.
In our application, the context-based affect analysis normally activates the personal context
modeling using the Bayesian networks first and then follows the emotional social context
modeling using ART-1 to justify or further derive the affect conveyed in the current input. For
example, in the above Homophobic bullying example transcript, the emotional context of Tiffany
is retrieved as ‘angry (1st input), disapproval (4th input) and angry (5th input)’. Thus we use the
Bayesian network first to predict the most likely affective state conveyed in Tiffany’s very last
input. The emotional state ‘angry’ has achieved the highest probability and been regarded as the
affect mostly likely implied in the input (“ONCE A BATTY ALWAYS A BATTY”). Since the most
relevant discussion context contributed by the 5th – 8th input is also sensed as being ‘negative’
using the ART-1 approach discussed above, we conclude that the very last input from Tiffany is
more likely to be ‘angry’ with a strong intensity indicated by the capitalization. Thus we adjust the
affect implied in the very last input from ‘neutral’ to ‘angry’.
In this way, we can predict the next most probable emotional state based on a character’s
previous emotional implications during the improvisation using the Bayesian networks and detect
the ‘positive or negative’ emotional implication of the most related discussion context using
unsupervised learning. The integration of both discussed approaches has great potential to derive
affect in communication context which is closer to the user’s real emotional experience. Another
advantage of our implementation is that it has the potential to perform contextual affect sensing
across different scenarios.
At the test stage, our affect detection component integrated with the AI agent detects affect for
each user input solely based on the analysis of individual turn-taking input itself as usual. The
above algorithms for context-based affect sensing will be activated when the affect detection
component recognizes ‘neutral’ from the current input during the emotionally charged proper
improvisation and the input also containing statement structures.
In this way, by considering the potential improvisational mood one character was in and recent
social emotional profiles of other characters, our affect detection component has been able to
inference emotion based on context to adjust the affect interpreted by the analysis based on
individual turn-taking user input. After the description of various affect processing components,
the overall affect detection model is shown in Figure 2.
12. Li Zhang
International Journal of Computational Linguistics (IJCL), Volume (1): Issue (4) 56
FIGURE 2: The Affect Detection Model
FIGURE 3: Affect Detection and the Control of Characters
The detected affective states from users’ open-ended text input play an important role in
producing emotional animation of human players’ avatars. The emotional animation mainly
includes emotional gesture and social attention (such as eye gazing). The expressive animation
engine, Demeanour [23], makes it possible for our characters to express the affective states
detected by the AI actor, EMMA. When it detects an affective state in a user’s text input, this is
passed to the Demeanour system attached to this user’s character and a suitable emotional
animation is produced. The Demeanour system has also used character profiles, particularly
including personality traits and relationships with other characters, to provide expressive
animation for other avatars when the ‘speaking’ avatar experiences affect. How the detected
affective states inform the animation engine and control the AI agent is illustrated in Figure 3.
Pre-processing
Affect detection based on
individual turn-taking
Affect justification based on
context profiles
The detected affect
One’s improvisational
mood, other characters’
recent emotional
profiles, discussion topic
etc
Affect detection from
metaphorical expression
(animal/food/affect
metaphors etc)
Affect detection from
literal expression
(imperatives etc)
13. Li Zhang
International Journal of Computational Linguistics (IJCL), Volume (1): Issue (4) 57
3. USER TESTING OF THE IMPROVISATIONAL AI ACTOR
We conducted an intensive user test with 160 secondary school students, in order to try out and
refine a testing methodology. The aim of the testing was primarily to measure the extent to which
having the AI agent as opposed to a person play a character affects users’ level of enjoyment,
sense of engagement, etc. We concealed the fact that the AI-controlled agent was involved in
some sessions in order to have a fair test of the difference that is made. We obtained surprisingly
good results. Having a minor bit-part character called “Dave” played by the AI agent as opposed
to a person made no statistically significant difference to measures of user engagement and
enjoyment, or indeed to user perceptions of the worth of the contributions made by the character
“Dave”. Users did comment in debriefing sessions on some utterances of Dave’s, so it was not
that there was a lack of effect simply because users did not notice Dave at all. Furthermore, it
surprised us that few users appeared to realize that sometimes Dave was computer-controlled.
We stress, however, that it is not an aim of our work to ensure that human actors do not realize
this.
Inspection of the transcripts collected indicates that the AI agent usefully pushed the
improvisation forward on various occasions. Figure 4 shows an example about how the AI actor
contributed to the drama improvisation in Crohn’s disease scenario. Briefly, in Crohn’s disease
scenario, Peter has had Crohn’s disease since the age of 15. Crohn’s disease attacks the wall of
the intestines and makes it very difficult to digest food properly. The character has the option to
undergo surgery (ileostomy) which will have a major impact on his life. The task of the role-play is
to discuss the pros and cons with friends and family and decide whether he should have the
operation. The other characters are; Mum, who wants Peter to have the operation, Matthew
(older brother) who is against the operation, Dad who is not able to face the situation, and David
(the best friend) who mediates the discussion. In the example transcript shown in Figure 4, Dave
was played by the AI actor, which successfully led the improvisation on the desirable track. In
another scenario, the Homophobic bullying, used for the testing, Mr. Dhanda was sometimes
played by EMMA and example transcripts are also shown that the AI actor has helped to push the
improvisation forward.
We have conducted an initial evaluation of the quality of the AI agent’s determinations about
emotion during these testing sessions, by comparing the AI agent’s determinations during one of
the Crohn’s disease improvisations with emotion labels later assigned offline by two members of
our team (not involved in the development of the AI agent’s algorithms). We used the kappa
statistic of Carletta [24]. It is a measure of the pairwise agreement among a set of coders making
category judgements, correcting for expected chance agreement. The statistic, K, is calculated as
K= (P(A)-P(E))/(1-P(E)) where P(A) is the proportion of times two coders agree and P(E) is the
proportion of times we would expect them to agree if they categorized randomly. A value of at
least 0.6 – 0.8 is generally required by researchers looking for good inter-annotator agreement.
We calculated K for each pair among the three labellers (EMMA and two humans). The inter-
human K was only 0.35, and so it is not surprising that the EMMA/human values were only 0.34
and 0.32. Although they are not ideal, at least these results give grounds for hope that our affect
detection with further refinement can come near the rather low human/human level of agreement.
Moreover, the overall accuracy rate of affect interpretation based on context in our current
analysis achieves 68% via the comparison of the annotation of part of the recorded transcripts
between human annotators and the AI agent in the Crohn’s disease scenario.
From the inspection of the evaluation results, although context affect detection is provided, there
are still some cases: when the two human judges both believed that user inputs carried negative
affective states (such as angry, threatening, disapproval etc), EMMA regarded them as neutral.
One most obvious reason is that some of the previous pipeline processing (such as dealing with
mis-spelling, acronyms etc, and syntactic processing from Rasp etc) failed to recover the
standard user input or recognize the complex structure of the input which led to less interesting
and less emotional context for some of the characters and may affect the performance of
contextual affect sensing. We also aim to extend the evaluation of the context-based affect
detection using transcripts from other scenarios. Also, since the test transcripts contained a very
14. Li Zhang
International Journal of Computational Linguistics (IJCL), Volume (1): Issue (4) 58
small number of metaphorical language phenomena comparatively, we intend to use other
resources (e.g. Wallstreet Journal and other metaphorical databases (e.g. ATT-Meta)) to further
evaluate the new development on metaphorical affect sensing.
FIGURE 4: Part of an Example Transcript Produced in the User Testing
The preliminary results from statistical analysis of the collected post questionnaires also indicate
that when the AI actor is involved in the improvisation, users’ abilities to concentrate on the
improvisation are somewhat higher in Crohn’s disease scenario than Homophobic bullying
scenario. When the AI actor is not involved in the improvisation, users’ abilities to concentrate on
the improvisation are a lot higher in Homophobic bullying than Crohn’s disease. This seems very
interesting, as it seems to be showing that the AI actor can make a real positive difference to an
aspect of user engagement when the improvisation is comparatively uninteresting.
4. CONCLUSIONS
Our work makes a contribution to the issue of what types of automation should be included in
interactive narrative environments, and as part of that the issue of what types of affect should be
detected (by directors, etc.) and how. Moreover, our work also makes a contribution to the
development of automatic understanding of human language and emotion. Our contextual affect
sensing shows initial directions for emotion modeling in personal and social context across
scenarios. Future work could include the equipment of the AI agent with the ability of performing
autonomous learning through metaphorical expressions.
5. REFERENCES
1. R.W. Picard. “Affective Computing”. The MIT Press. Cambridge MA. 2000
2. A. Ortony, G.L. Clore & A. Collins. “The Cognitive Structure of Emotions”. Cambridge U.
Press. 1998
3. A. Egges, S. Kshirsagar & N. Magnenat-Thalmann. “A Model for Personality and Emotion
Simulation”, In Proceedings of Knowledge-Based Intelligent Information & Engineering
Systems (KES2003), Lecture Notes in AI. Springer-Verlag: Berlin. 2003
4. R.S. Aylett, J. Dias and A. Paiva. “An affectively-driven planner for synthetic characters”. In
Proceedings of ICAPS. 2006
5. Nogueiras et al. “Speech emotion recognition using hidden Markov models”. In Proceedings of
Eurospeech 2001, Denmark. 2001
15. Li Zhang
International Journal of Computational Linguistics (IJCL), Volume (1): Issue (4) 59
6. M. Pantic, A. Pentland, A. Nijholt and T. Huang. “Human Computing and Machine
Understanding of Human Behavior: A Survey”. In Proc. Int’l Conf. Multimodal Interfaces, pp.
239-248. 2006
7. M. Mateas. Ph.D. Thesis. “Interactive Drama, Art and Artificial Intelligence”. School of
Computer Science, Carnegie Mellon University. 2002
8. X. Zhe & A.C. Boucouvalas. “Text-to-Emotion Engine for Real Time Internet Communication”.
In Proceedings of International Symposium on Communication Systems, Networks and DSPs,
Staffordshire University, UK, pp 164-168. 2002
9. D. Watson & A. Tellegen. “Toward a Consensual Structure of Mood”. Psychological Bulletin,
98, 219-235. 1985
10. P. Ekman. “An Argument for Basic Emotions”. In Cognition and Emotion, 6, 169-200. 1992
11. L. Zhang, J.A. Barnden, R.J. Hendley & A.M. Wallington. “Developments in Affect Detection in
E-drama”. In Proceedings of EACL 2006, 11th Conference of the European Chapter of the
Association for Computational Linguistics, 2006, Trento, Italy. pp. 203-206. 2006
12. E. Briscoe and J. Carroll. “Robust Accurate Statistical Annotation of General Text”. In
Proceedings of the 3rd International Conference on Language Resources and Evaluation, Las
Palmas, Gran Canaria. 1499-1504. 2002
13. D.R. Heise. “Semantic Differential Profiles for 1,000 Most Frequent English Words”.
Psychological Monographs. 70 8:(Whole 601). 1965
14. S. Fussell & M. Moss. “Figurative Language in Descriptions of Emotional States”. In S. R.
Fussell and R. J. Kreuz (Eds.), Social and cognitive approaches to interpersonal
communication. Lawrence Erlbaum. 1998
15. Z. Kövecses. “Are There Any Emotion-Specific Metaphors?” In Speaking of Emotions:
Conceptualization and Expression. Athanasiadou, A. and Tabakowska, E. (eds.), Berlin and
New York: Mouton de Gruyter, 127-151. 1998
16. J. Barnden, S. Glasbey, M. Lee & A. Wallington. “Varieties and Directions of Inter-Domain
Influence in Metaphor”. Metaphor and Symbol, 19(1), 1-30. 2004
17. J.A. Barnden. “Metaphor, Semantic Preferences and Context-sensitivity”. Invited chapter for
a Festschrifft volume. Kluwer. 2006
18. C. Strapparava and A. Valitutti. “WordNet-Affect: An Affective Extension of WordNet”, In
Proceedings of the 4th International Conference on Language Resources and Evaluation
(LREC 2004), Lisbon, Portugal, 1083-1086. 2004
19. P. Rayson. “Matrix: A statistical method and software tool for linguistic analysis through
corpus comparison”. Ph.D. thesis, Lancaster University. 2003
20. C. Fellbaum. “WordNet, an Electronic Lexical Database”. The MIT press. 1998
21. D. Sperber & D. Wilson. “Relevance: Communication and cognition (2nd ed.)”. Oxford, UK:
Blackwell. 1995.
22. D. Wilson & D. Sperber. “Relevance Theory”. In G.Ward & L. Horn (Eds.), Handbook of
Pragmatics (pp. 607–632). Oxford, UK: Blackwell. 2003.
16. Li Zhang
International Journal of Computational Linguistics (IJCL), Volume (1): Issue (4) 60
23. M. Gillies, I.B. Crabtree and D. Ballin. “Individuality and Contextual Variation of Character
Behaviour for Interactive Narrative”. In Proceedings of the AISB Workshop on Narrative AI and
Games. 2006
24. J. Carletta. “Assessing Agreement on Classification Tasks: The Kappa statistic.”
Computational Linguistics, 22 (2), pp.249-254. 1996
25. L. Zhang, M. Gillies, K. Dhaliwal, A. Gower, D. Robertson & B. Crabtree. “E-drama:
Facilitating Online Role-play using an AI Actor and Emotionally Expressive Characters”.
International Journal of Artificial Intelligence in Education. Vol 19(1), pp.5-38. 2009
26. J.M. Lopez, R. Gil, R., Garcia, I. Cearreta and N. Garay. “Towards an Ontology for Describing
Emotions”. In WSKS '08 Proceedings of the 1st world summit on The Knowledge Society:
Emerging Technologies and Information Systems for the Knowledge Society. 2008