El documento describe una aplicación de chat gratuita llamada LIFE que permite a los usuarios comunicarse con personas de todo el mundo. La aplicación traduce mensajes entre más de 15 idiomas con alta precisión. Se destaca que la aplicación se ve y se siente similar a iPhone, lo que la hace atractiva. Además, es una buena opción para aquellos que quieren practicar idiomas extranjeros o hacer amigos internacionales de manera fácil.
LIFE es una nueva aplicación de mensajería social similar a LINE y Twitter. Permite chatear personalmente y en grupos de forma gratuita, al igual que LINE. También permite comunicarse con personas de todo el mundo de forma anónima como en Twitter. Una ventaja única de LIFE es que los usuarios pueden ver si sus mensajes han sido leídos y pueden reaccionar a los mensajes de otros en los chats grupales.
Testudines, ou quelônios, são répteis caracterizados por possuírem um casco formado por uma carapaça dorsal e um plastrão ventral. Existem 327 espécies de Testudines no mundo, incluindo tartarugas marinhas, terrestres e de água doce. No Rio Grande do Sul ocorrem 11 espécies nativas, como o cágado-preto e a tartaruga-tigre-d'água.
(1) A pilot study in the Netherlands tested having an occupational physician (OP) work as a consultant in 7 primary health care centers to help general practitioners (GPs) with patient referrals, especially for work-related health issues.
(2) Over 7 months, the OPC consulted on 184 patient referrals from GPs, with the most common diagnoses being psychological (52%) and musculoskeletal (26%).
(3) Based on the results, GPs found the OPC useful in identifying work-related health problems early to provide appropriate care and advice on returning to work.
Terence Donovan foi um fotógrafo britânico conhecido por sua fotografia de moda na década de 1960. Ele estudou fotografia entre os 11 e 15 anos e trabalhou com a Vogue de 1963 até sua morte em 1996. Seu estilo informal e imagens em preto e branco o distinguia por retratar modelos em ambientes urbanos e rústicos.
O documento descreve momentos importantes no desenvolvimento da fotografia ambiental entre 1827 e 1860, incluindo a primeira fotografia permanente criada por Joseph Nicéphore Niépce em 1827, a primeira foto de uma pulga por John Benjamin Dancer em 1840, e o primeiro livro ilustrado com fotografias, Photographs of British Algae de Anna Atkins, em 1843.
Hays Journal 11 - Wereldwijd Inzicht voor Experts in de Wereld van WerkHays Netherlands
Hays Journal is een tweejaarlijkse publicatie met inzichten en nieuws voor HR, recruitment professionals en HR-managers over de steeds veranderende wereld van werk.
Bekijk de uitgave online op https://www.hays.nl/hays-journal/index.htm of vraag een print exemplaar aan bij marcom@hays.nl.
El documento describe una aplicación de chat gratuita llamada LIFE que permite a los usuarios comunicarse con personas de todo el mundo. La aplicación traduce mensajes entre más de 15 idiomas con alta precisión. Se destaca que la aplicación se ve y se siente similar a iPhone, lo que la hace atractiva. Además, es una buena opción para aquellos que quieren practicar idiomas extranjeros o hacer amigos internacionales de manera fácil.
LIFE es una nueva aplicación de mensajería social similar a LINE y Twitter. Permite chatear personalmente y en grupos de forma gratuita, al igual que LINE. También permite comunicarse con personas de todo el mundo de forma anónima como en Twitter. Una ventaja única de LIFE es que los usuarios pueden ver si sus mensajes han sido leídos y pueden reaccionar a los mensajes de otros en los chats grupales.
Testudines, ou quelônios, são répteis caracterizados por possuírem um casco formado por uma carapaça dorsal e um plastrão ventral. Existem 327 espécies de Testudines no mundo, incluindo tartarugas marinhas, terrestres e de água doce. No Rio Grande do Sul ocorrem 11 espécies nativas, como o cágado-preto e a tartaruga-tigre-d'água.
(1) A pilot study in the Netherlands tested having an occupational physician (OP) work as a consultant in 7 primary health care centers to help general practitioners (GPs) with patient referrals, especially for work-related health issues.
(2) Over 7 months, the OPC consulted on 184 patient referrals from GPs, with the most common diagnoses being psychological (52%) and musculoskeletal (26%).
(3) Based on the results, GPs found the OPC useful in identifying work-related health problems early to provide appropriate care and advice on returning to work.
Terence Donovan foi um fotógrafo britânico conhecido por sua fotografia de moda na década de 1960. Ele estudou fotografia entre os 11 e 15 anos e trabalhou com a Vogue de 1963 até sua morte em 1996. Seu estilo informal e imagens em preto e branco o distinguia por retratar modelos em ambientes urbanos e rústicos.
O documento descreve momentos importantes no desenvolvimento da fotografia ambiental entre 1827 e 1860, incluindo a primeira fotografia permanente criada por Joseph Nicéphore Niépce em 1827, a primeira foto de uma pulga por John Benjamin Dancer em 1840, e o primeiro livro ilustrado com fotografias, Photographs of British Algae de Anna Atkins, em 1843.
Hays Journal 11 - Wereldwijd Inzicht voor Experts in de Wereld van WerkHays Netherlands
Hays Journal is een tweejaarlijkse publicatie met inzichten en nieuws voor HR, recruitment professionals en HR-managers over de steeds veranderende wereld van werk.
Bekijk de uitgave online op https://www.hays.nl/hays-journal/index.htm of vraag een print exemplaar aan bij marcom@hays.nl.
Joseph Nicéphore Niépce cria a primeira fotografia permanente em 1827. John Benjamin Dancer fotografa uma pulga usando um microscópio em 1840. Anna Atkins publica o primeiro livro ilustrado com fotografias, Photographs of British Algae, em 1843.
O documento descreve diferentes tipos de fotografia científica, incluindo eletromicrografia, microscopia eletrônica de varredura, fotografia de alta velocidade, fotografia de fluorescência e fotografia documental. Ele fornece detalhes sobre como cada técnica é usada e quais informações podem ser obtidas através dela.
On this presentation we will have an opportunity to see how can we predict churn in telco industry. Marko will tell us how he done this at Telenor Srbija.
Apache Spark 2.0: Faster, Easier, and SmarterDatabricks
In this webcast, Reynold Xin from Databricks will be speaking about Apache Spark's new 2.0 major release.
The major themes for Spark 2.0 are:
- Unified APIs: Emphasis on building up higher level APIs including the merging of DataFrame and Dataset APIs
- Structured Streaming: Simplify streaming by building continuous applications on top of DataFrames allow us to unify streaming, interactive, and batch queries.
- Tungsten Phase 2: Speed up Apache Spark by 10X
O documento descreve duas figuras importantes da fotografia: Lucien Cleguer, fotógrafo francês que promoveu grandes exposições de fotografia na Europa e foi amigo de Picasso, e Schwab Jo Born, fotógrafo alemão de moda que se concentra em retratos para revistas e publicidade.
Filip Panjevic is a Co-Founder and CTO at ydrive.ai - startup dealing with self-driving cars, and one of the founders of Petnica Machine Learning School.
Filip's talk will focus on the story of Petnica School, how did it start, what has changed since the beginning, how the concept of school looks right now and why is that concept good for making new data scientists. This talk will be perfect for people who consider starting their careers in the data science field!
The talk will be a broad overview and thoughts about building one of the biggest data science communities in India. I will talk about how an ecosystem is created and value delivered to each stakeholder. I will be sharing my experience of building MachineHack and AIMinds and other platforms. One of the core agendas of the talk will be how these platforms have enabled a unique data science education and learning experience in India. The platforms built help students and engineers to imagine and work towards a career in data science.
In Drazen talk, you will get a chance to listen to how Data Science Master 4.0 on Belgrade University was created, and what are the benefits of the program.
PwC's recently released Responsible AI Diagnostic surveyed around 250 senior business executives from May to June 2019. The survey says that 84% of CEOs agree that AI-based decisions need to be explainable in order to be trusted. In the past few years, Deep learning has shown remarkable results in various applications, which makes it one of the first choices for many AI use cases. However, deep learning models are hard to explain, and since the majority of CEOs expect AI solutions to be explainable, deep learning has a serious challenge. Daniel Kahneman, in his book thinking fast and slow, presented two different systems the human brain uses to form thoughts and decisions: System 1: fast, intuitive and hard to explain System 2: slow, conscious and easy to explain In this talk I will present: A) PwC Responsible AI Survey B) A proposed deep learning framework that mimics the two systems of thinking C) The recent advances in the neural symbolic learning field.
Challenges in building a churn prediction model in different industries, presented by Jelena Pekez from Comtrade System Integration. Talk is focused on real-life use-case experience.
This document discusses using business intelligence (BI) to improve risk management at a bank. It provides three key ways BI can create value: protecting revenue, improving risk assessments, and reducing operational costs. Specific use cases are described, including early warning systems, behavioral detection systems, and modern BI platforms that combine data aggregation, analytics, and infrastructure for faster insights. The presentation outlines a proof of concept and roadmap for implementing a modern BI system at the bank to enable self-service analytics, alerts, automated data delivery, and collaboration across the organization. Dashboards and data insights are shown as examples of the types of risk analyses and reporting that will be possible with a new modern BI platform.
The talk will have 3 parts. The overview of the practical applications of the AI and ML in the FinTech industry with a short explanation of the PSD2 directive and the disruption is caused. Application of the AI/ML from the perspective of the end-user, personal financial health, financial coach, etc. The overview of the architecture, technologies, and frameworks used with practical examples from the Zuper company.
We present a recommender system for personalized financial advice, which we designed for a large Swiss private bank. The final recommendations produced by the system were delivered to the end clients through a mobile banking platform. The recommender system is based on a collaborative filtering technique and can work with changing asset features, operate with implicit ratings and react to explicit feedback that clients can give using the mobile app. Moreover, we developed and implemented an approach to provide an explanation for each recommendation in the form “As you bought A, you might like B".
This talk shall focus on making real-time pipelines using cutting edge Big Data technologies and applying ML on gathered data. The first part of the presentation shall cover importance and necessity for streaming data processing. In addition, tools that could be used in order to build a streaming pipeline shall be proposed. The second part of this talk shall focus on making machine learning models in customer support. There shall be introduced success stories covering the need for more efficient customer support, problem resolution and gained benefits.
Presentation of the first complete AI investment platform. It is based on most innovative AI methods: most advanced neural networks (ResNet/DenseNet, LSTM, GAN autoencoders) and reinforcement learning for risk control and position sizing using Alpha Zero approach. It shows how the complex AI system which covers both supervised and reinforcement learning could be successfully used to investment portfolio optimization in real-time. The architecture of the platform and used algorithms will be presented together with the workflow of machine learning. Also, the real demo of the platform will be shown.
A lot of companies make the mistake of thinking that just hiring Data Scientists will lead to increased revenue or increased profit. For a company’s investment in Data Science to be successful the Data Scientists need to work on the right problems, with the right people, and with the right tools. In this presentation, I will talk about the lessons I have learned, and mistakes made in applying Data Science in commercial settings over the last 10 years. I will highlight what processes can increase the chances of Data Science investment being successful.
Joseph Nicéphore Niépce cria a primeira fotografia permanente em 1827. John Benjamin Dancer fotografa uma pulga usando um microscópio em 1840. Anna Atkins publica o primeiro livro ilustrado com fotografias, Photographs of British Algae, em 1843.
O documento descreve diferentes tipos de fotografia científica, incluindo eletromicrografia, microscopia eletrônica de varredura, fotografia de alta velocidade, fotografia de fluorescência e fotografia documental. Ele fornece detalhes sobre como cada técnica é usada e quais informações podem ser obtidas através dela.
On this presentation we will have an opportunity to see how can we predict churn in telco industry. Marko will tell us how he done this at Telenor Srbija.
Apache Spark 2.0: Faster, Easier, and SmarterDatabricks
In this webcast, Reynold Xin from Databricks will be speaking about Apache Spark's new 2.0 major release.
The major themes for Spark 2.0 are:
- Unified APIs: Emphasis on building up higher level APIs including the merging of DataFrame and Dataset APIs
- Structured Streaming: Simplify streaming by building continuous applications on top of DataFrames allow us to unify streaming, interactive, and batch queries.
- Tungsten Phase 2: Speed up Apache Spark by 10X
O documento descreve duas figuras importantes da fotografia: Lucien Cleguer, fotógrafo francês que promoveu grandes exposições de fotografia na Europa e foi amigo de Picasso, e Schwab Jo Born, fotógrafo alemão de moda que se concentra em retratos para revistas e publicidade.
Filip Panjevic is a Co-Founder and CTO at ydrive.ai - startup dealing with self-driving cars, and one of the founders of Petnica Machine Learning School.
Filip's talk will focus on the story of Petnica School, how did it start, what has changed since the beginning, how the concept of school looks right now and why is that concept good for making new data scientists. This talk will be perfect for people who consider starting their careers in the data science field!
The talk will be a broad overview and thoughts about building one of the biggest data science communities in India. I will talk about how an ecosystem is created and value delivered to each stakeholder. I will be sharing my experience of building MachineHack and AIMinds and other platforms. One of the core agendas of the talk will be how these platforms have enabled a unique data science education and learning experience in India. The platforms built help students and engineers to imagine and work towards a career in data science.
In Drazen talk, you will get a chance to listen to how Data Science Master 4.0 on Belgrade University was created, and what are the benefits of the program.
PwC's recently released Responsible AI Diagnostic surveyed around 250 senior business executives from May to June 2019. The survey says that 84% of CEOs agree that AI-based decisions need to be explainable in order to be trusted. In the past few years, Deep learning has shown remarkable results in various applications, which makes it one of the first choices for many AI use cases. However, deep learning models are hard to explain, and since the majority of CEOs expect AI solutions to be explainable, deep learning has a serious challenge. Daniel Kahneman, in his book thinking fast and slow, presented two different systems the human brain uses to form thoughts and decisions: System 1: fast, intuitive and hard to explain System 2: slow, conscious and easy to explain In this talk I will present: A) PwC Responsible AI Survey B) A proposed deep learning framework that mimics the two systems of thinking C) The recent advances in the neural symbolic learning field.
Challenges in building a churn prediction model in different industries, presented by Jelena Pekez from Comtrade System Integration. Talk is focused on real-life use-case experience.
This document discusses using business intelligence (BI) to improve risk management at a bank. It provides three key ways BI can create value: protecting revenue, improving risk assessments, and reducing operational costs. Specific use cases are described, including early warning systems, behavioral detection systems, and modern BI platforms that combine data aggregation, analytics, and infrastructure for faster insights. The presentation outlines a proof of concept and roadmap for implementing a modern BI system at the bank to enable self-service analytics, alerts, automated data delivery, and collaboration across the organization. Dashboards and data insights are shown as examples of the types of risk analyses and reporting that will be possible with a new modern BI platform.
The talk will have 3 parts. The overview of the practical applications of the AI and ML in the FinTech industry with a short explanation of the PSD2 directive and the disruption is caused. Application of the AI/ML from the perspective of the end-user, personal financial health, financial coach, etc. The overview of the architecture, technologies, and frameworks used with practical examples from the Zuper company.
We present a recommender system for personalized financial advice, which we designed for a large Swiss private bank. The final recommendations produced by the system were delivered to the end clients through a mobile banking platform. The recommender system is based on a collaborative filtering technique and can work with changing asset features, operate with implicit ratings and react to explicit feedback that clients can give using the mobile app. Moreover, we developed and implemented an approach to provide an explanation for each recommendation in the form “As you bought A, you might like B".
This talk shall focus on making real-time pipelines using cutting edge Big Data technologies and applying ML on gathered data. The first part of the presentation shall cover importance and necessity for streaming data processing. In addition, tools that could be used in order to build a streaming pipeline shall be proposed. The second part of this talk shall focus on making machine learning models in customer support. There shall be introduced success stories covering the need for more efficient customer support, problem resolution and gained benefits.
Presentation of the first complete AI investment platform. It is based on most innovative AI methods: most advanced neural networks (ResNet/DenseNet, LSTM, GAN autoencoders) and reinforcement learning for risk control and position sizing using Alpha Zero approach. It shows how the complex AI system which covers both supervised and reinforcement learning could be successfully used to investment portfolio optimization in real-time. The architecture of the platform and used algorithms will be presented together with the workflow of machine learning. Also, the real demo of the platform will be shown.
A lot of companies make the mistake of thinking that just hiring Data Scientists will lead to increased revenue or increased profit. For a company’s investment in Data Science to be successful the Data Scientists need to work on the right problems, with the right people, and with the right tools. In this presentation, I will talk about the lessons I have learned, and mistakes made in applying Data Science in commercial settings over the last 10 years. I will highlight what processes can increase the chances of Data Science investment being successful.
The talk would be focusing on reasons and method for creating models which maximize sales price Gross Margin but still has high confidentiality that quote would be accepted by the client. Price changes are dynamic things that are impacted by many different elements like cost of input material, labor cost, transportation cost, scrap material due to different ordered quantities, etc. Besides input cost segments, output price is also impacted by different marketing campaigns (own and others), seasonality, past and future customer behavior as well as the behavior of the product we are selling.
Data is now a valuable asset for businesses, as companies that effectively use data are outperforming their peers by moving further ahead faster and more cost-effectively. However, some businesses remain indifferent to the value of data, failing to take charge of this new gold, while customer expectations have never been higher. Coeus claims to add fuel to businesses by providing higher conversion rates through effective use of data.
In the past few years, many businesses started do understand the potential of real-time data analytics. And many of those invested time, energy and finances to make it happen, with weaker outcomes than expected. Reasons are few for this: too ambitious plans by leadership regarding leveraging data, not enough discipline defining goals and MVP for initial use cases, a plethora of tools and vendors available who claim that can solve all the problems, etc. So, how can we get the most value with reasonable costs out of fast (real-time) data? We will try to answer this question and give actionable advice.
This document discusses the design of a personalized 3A health monitoring system using sensor networks. It begins with an overview of current challenges in healthcare like an aging population and increasing costs. It then describes the proposed system which would use sensors, edge computing, blockchain and other technologies to provide continuous remote health monitoring anywhere and anytime. Key aspects of the system include a heart rate monitoring solution, data exchange centers, smart health homes and eHealth labs. The system aims to address issues like data ownership and security while providing personalized care. It concludes by discussing next steps to test and implement the continuous monitoring system.
This document discusses improving data quality through product similarity search. It describes leveraging multiple data sources like product names, descriptions, prices and attributes to calculate similarity between products. Different techniques are used depending on the data type, such as text similarity for names/descriptions, numerical similarity for prices and variant counts, and mixed similarity for attributes. Attributes require special handling due to different data types within. The document outlines challenges in comparing incompatible datasets and noisy data. It proposes a solution using an API that can customize similarity based on use cases and data specificities.
Uroš Valant has almost 20 years of experience in planning, managing and delivering of various IT projects. He has the best and richest experience in the field of business analytics, project planning and implementation, database design and the management of development teams. In the last years, his focus is the field of predictive analytics, machine learning and applying the AI solution to a practical use in different field of work.
In his talk he will present to us interactive case study of the image recognition use and AI assisted design techniques in the textile industry.
The presentation will start as an engaging lecture where I will present the motivation behind the project based on my academic research (my Oxford PhD among others). I will tell the audience just how rampant corruption is in local governance and why is it so persistent. Then I will present our remedy: full budget transparency. I will show them our search engine and how it works, and will call the participants to download the APIs and play with the data themselves.
The talk will be divided into two parts. The first one is about geospatial open data and several Copernicus services where those data can be downloaded. The second one is about Forest and Climate project, as an example of geospatial analysis. The aim of the project was to identify the most suitable area for afforestation in Serbia by using satellite and Earth observation data. The results can be found at https://sumeiklima.org/.
Geospatial Analysis and Open Data - Forest and Climate
Application of text mining and graph database on civil engineering projects - Djordje Nedeljkovic
1. Đorđe Nedeljković,
Faculty of Civil Engineering, teaching assistant
Department of construction project management
Application of Text Mining and graph
database on civil engineering projects
2. Predmet istraživanja
- Pretraga, izdvajanje, analiziranje i vizuelizacija znanja iz
nestruktuiranih/polustruktuiranih dokumenata sa građevinskih
projekata
- Osnovni zadatak (klasifikacija) – dobri rezultati sa BoW modelom
- Kompleksniji zadaci – loši rezultati, potreban novi feature vector
3. Sadržaj
- Građevinski projekti / tehnički dokumenti
- Postojeća rešenja
- Predloženi model
- Pretpostavke
- Mere asocijacije
- Reprezentacija detektovanih ključnih fraza
- Povezivanje značajnih fraza na osnovu semantičke bliskosti
- Pravila za izdvajanje koncepata i relacija
- Primeri
- Zaključak
4. Investicioni projekat (građevinski, arhitektonski)
- Za razliku od projekata u opštem smislu, izgradnja,
rekonstrukcija, modifikacija i opremanje investicionih objekata
su uvek u direktnoj vezi sa građevinarstvom kao privrednom
granom
- Kompleksan tehničko-tehnološki, organizacioni, finansijski i
pravni poduhvat, koji se sastoji od skupa koordinisanih i
kontrolisanih aktivnosti sa jasno definisanim početkom i
krajem, čiji je cilj izgradnja, rekonstrukcija, modifikacija i/ili
opremanje objekta ili objekata koji su potrebni vlasniku
(investitoru)
5. Investicioni projekat - specifičnosti
- Složen
- Unikatan
- Na više lokacija
- Dugotrajan
- Veliki broj učesnika
- Razuđenost procesa
- Važnost klimatskih uslova
- Imovinsko-pravni problemi
6. Dokumenti na građevinskom projektu
- Pored tehničkih crteža i proračuna, značajan korpus tekstualnih
dokumenata, (posebno u fazi realizacije projekta):
Zapisnici sa sastanaka, varijacije, klejmovi, fakture, izveštaji, dopisi...
- Veliki broj učesnika sa različitim poslovnim procesima i
stepenom ICT zrelosti
Mane Prednosti
Statički, neinformativni sadržaj
(zaglavlja, formulari, itd.)
Konzistentna struktura
Domenski žargon, skraćenice
Manje višeznačnih jezičkih konstrukcija
(polisemija, metonimija, itd.)
Dužina, više tema
Sadržaj na različitim jezicima
(često na nivou rečenice)
7. Trendovi na domaćem tržištu
- Alati koji se najčešće koriste za obradu podataka:
Programi za rad sa tabelama
- Prepreke za prelazak na napredniji alat za obradu podataka:
Nekompatibilnost sa postojećim poslovnim procesima
Podaci su u neodgovarajućem formatu za pretragu i analizu
- Prepreke za optimalno korišćenje nestruktuiranih podataka u
procesu donošenja odluka:
Značajni podaci se nalaze na različitim mestima
8. Postojeća rešenja za pretragu,
izdvajanje, analiziranje i vizuelizaciju
- Ručno obeležavanje, rad sa prethodno definisanim formama
- Information exctraction, Ontology based, Semantic annotation
- Document management system
- Enterprise search
- BI applications
- Sales enablement software
- Content management system
- Enterprise resource planning
9. Predloženi model - hipoteze
- Robusnost na nedostatak NLP resursa, podrška za više jezika
Izdvajanje ključnih fraza zasnovano na merama asocijacije reči
- Transferabilnost na različite domene sa minimalnim trudom
eksperta za konfiguraciju sistema prethodnim znanjem
Mogućnost definisanja prethodnog znanja kroz resurs fajlove i
zadata pravila
10. Inicijalno izdvajanje značajnih fraza
- Značajne fraze (ZF) kao par susednih reči
- Informativnije od pojedinačnih reči
- Mere za određivanje verovatnoće zajedničkog pojavljivanja
reči x i y u paru (x,y)
- Isti par može biti drugačije rangiran za različite mere
- Pojedinačno, mere preferiraju parove reči sa određenim
kombinacijama frekvencija
- Kombinovanje najbolje rangiranih parova za različite mere
12. Natural language processing (NLP) resursi
- Detektor jezika
- Nivo rečenice, zasnovan na frekvenciji najčešćih bigrama
- Lemmatizer
- Svođenje reči na kanonski oblik, kompaktniji rečnik
- Part-of-speach tagger
- Klasifikacija reči, dozvoljene kombinacije
13. Redukcija neinformativnog sadržaja
- Uvećan skor za parove reči u neinformativnim delovima teksta
- Česti parovi reči u istom kontekstu – šum
- Parovi reči u različitim kontekstima – informativne ZF
- Informativnost para reči – entropija skupa string reprezentacije svih
pojavljivanja
- Korigovanje skora dobijenog merama asocijacije
14. Uspostavljanje relacija
- Domenski nezavisan pristup – relacije između ZF na osnovu
kontekstualne sličnosti
- Mera - Jaccard indeks za skupove paragrafa/rečenica
- Relacije with, always_under, always_with
|Pki
∩ Pkj
|
| Pki
∪ Pkj
|
≥ t ∈ 0,1
- Grupisanje ZF od dve reči povezanih always_with relacijom
(Bron-Kerbosch algoritam)
Pki
Pkj
Pki
Pkj
Pki
Pkj
always_with
15. Graf značajnih fraza/dokumenata
- Značajne_fraze i dokumenti
kao čvorovi grafa
- Automatski generisan
- Parametari ekstrakcije
definišu strukturu
16. Definisanje dodatnih koncepata i relacija
- Automatska detekcija obrazaca u tekstu
Regularni izrazi - datum, novac
- Struktura dokumenata:
Zapisnici sa sastanaka - osoba, kompanija
Predmet i predračun - pozicija_rada
Work breakdown structure, gantogram – aktivnost, faza
- ZF koje ispunjavaju zadati uslov
konstruktivni_elementi, materijali
- Relacije:
osoba radi_za kompanija sastanak održan datum osoba akcija ZF
24. Zaključak (SW)
• Nezavisnost u odnosu na jezik
• Izdvajanje ključnih fraza na bazi entropije
• Analiza i vizuelizacija moguća bez prethodno
definisane reprezentacije domenskog znanja
• Nema potrebe za menjanjem postojećih
poslovnih procesa
• Integracija fragmentisanih podataka na nivou
dokumenta
• Lošija performansa bez NLP komponenti
• Novi pristup u domenu upravljanja
građevinskim projektima
• Razdvojeni rečnici značajnih fraza za različite
jezike
25. Zaključak (OT)
• Veliki udeo nestruktuiranih tekstualnih
podataka u projektnoj dokumentaciji
• Postprojektna analiza
• Upotreba ML tehnika za poboljšanje
performansi i nove funkcionalnosti
(klasifikacija, klastering, detekcija događaja,...)
• Network analysis
(SNA, Link analysis, Centrality measures)
• Teškoća da se proceni uticaj na proces
donošenja odluka
• Prava pristupa nad informacijama iz više
dokumenata
• Definisanje pravila za izdvajanje novog znanja
Dobro jutro svima, moje ime je Djordje Nedeljkovic, student doktorskih studija I asistent na gradjevinskom fakultetu Univerziteta u Beogradu.
Pre nego sto formalno zapocnem sa pricom, mala digresija – kazu da je dobra praksa da se autor potrudi i prezentaciju prozme sa malo duhovitih momenata kako bi se ljudi opustili I razbila monotonija.
Ja necu morati da se trudim oko toga jer nosim fiksnu protezu koja ce sama pobrinuti da bude saljivih momenata tokom prezentacije. Ovo je ujedno I izvinjenje za neke reci koje ce možda iskociti, posebno ako imaju vise suglasnika.
Na prvi pogled je nelogicno sto se u naslovu ne pominje koji problem se resava, nego se navodi sta se koristi da bi se postigao cilj.
A taj cilj na kome radi tim sa gradj fakulteta na celu sa prof MK je da se iskoristi velika kolicina znanje koje je za sada uslovno receno skriveno u dokumentima sa gradj projekta.
Formalno, cilj je da se definise model koji ce omoguciti efikasnu pretragu, izdvajanje, analizu I vizuelizaciju znanja iz nestruk I polustruk dok sa gradj projekta.
Medjutim, ono sto je u ovoj formulaciji moze da bude problematicno I zbog cega ona nije naslov prezentacije je rec znanje.
Naime, postoje razlicite definicije, razliciti pragovi koji se moraju preskociti da bi neko mogao da tvrdi da operise na nivou znanja.
Posto ima sasvim dovoljno tema o kojima moze da se prica I pre nivoa znanja, danas se fokusiramo na motive istrazivanja, na samu prirodu gradj dokumentacije, na tehnike koje su koriscene da se nestruktuirani sadrzaj delimicno uredi I na koriscenje gbp za reprezentaciju dobijenih rezultata.
Mali uvod - kako je sve pocelo I sta je bio inicijalni motiv za istrazivanje? S obzirom se na katedri bavimo primenom data mininga I masinskim ucenjem u domenu gradjevinarstva, pre nekih god I po dana kolege sa fakulteta su nam dostavile korpus dokumenata vezanih za izmene ugovorenih radova na jednom kompleksnom medjunarodnom projektu.
Danas ce biti izlozen rezultati tog istrazivanja
Prvo ce biti prikazane karaktersitike gradj projekata i dokumentacije u domenu gradj industrije. Zatim sledi pregled postojecih pristupa za izdvajanje i analizu znanja iz tekstualnih izvora.
Kada se ogranicenja I potrebe korisnika na trzistu, moze se objasniti struktura predlozenog modela.
Na kraju ce biti prikazani primeri konkretnih rezultata koji bi se dobili u radu sa ovakvom reprezentacijom, kao I diskusija o predlozenom resenju.
Kada se govori o investicionim projektima, podrazumeva se realizacija niza aktivnosti, od same ideje o određenom investicionom dobru, preko izrade predinvesticionih studija, planske i projektne dokumentacije do ugovaranja, izgradnje, opremanja, obuke kadrova i puštanja objekta u eksploataciju. Navedeni spisak aktivnosti je samo jedan deo života investicionog objekta. Nije redak slučaj da se dopunjuje i aktivnostima koje su van tradicionalnih okvira definisanih za oblast građevinarstva, kao što su, na primer, aktivnosti na obezbeđenju neophodnih finansija za izgradnju i opremanje, transfer tehnologije koja se koristi u objektu, uspostavljanje raznovrsnih trgovinskih aranžmana, itd.
složenost – veliki broj raznovrsnih aktivnosti tokom realizacije projekta
unikatnost – svaki građevinski proizvod je unikatan, a to uzrokuje nemogućnost formiranja jedinstvenih cena
na jednoj ili više lokacija, ponekad značajno međusobno udaljenih, istovremeno funkcioniše više proizvodnih linija, na kojima se izvode raznovrsne aktivnosti
angažovan je veliki broj učesnika, od firmi, pojedinaca, do pojedinih gradskih i državnih ustanova i komunalnih preduzeća
razuđenost procesa – pokretač i finansijer posla - investitor može da razdvoji fazu projektovanja od faze izvođenja budućeg objekta, Time se posao na realizaciji projekta dodatno komplikuje, a broj učesnika povećava
značajan deo aktivnosti ostvaruje se na otvorenom prostoru, podložnost klimatskim uticajima – sezonski karakter
objekti se često grade u urbanizovanim područjima, zato se ističe važnost pravovremenog i potpunog rešavanja imovinsko-pravnih problema, obezbeđivanja uslova za projektovanje od komunalnih preduzeća i inspekcijskih organa, dobijanja saglasnosti na projektovanu dokumentaciju, itd.