A Graph-based Cross-lingual Projection Approach for Weakly Supervised Relation Extraction

Seokhwan Kim
Seokhwan KimScientist at Institute for Infocomm Research
A GRAPH-BASED CROSS-LINGUAL
      PROJECTION APPROACH FOR
WEAKLY SUPERVISED RELATION EXTRACTION
    The 50th Annual Meeting of the Association for Computational Linguistics
                                  (ACL 2012)
                             July 11th, 2012, Jeju

       Seokhwan Kim (Institute for Infocomm Research)
                 Gary Geunbae Lee (POSTECH)
Contents
• Introduction
• Methods
    Cross-lingual Annotation Projection for Relation Extraction
    Graph-based Projection Approach
• Evaluation
• Conclusions




                                                                   2
Contents
• Introduction
• Methods
    Cross-lingual Annotation Projection for Relation Extraction
    Graph-based Projection Approach
• Evaluation
• Conclusions




                                                                   3
Problem Definition
• Relation Extraction
    To identify semantic relations between a pair of entities

                         Birthplace

       Barack Obama was born in Honolulu           ,   Hawaii    .
              PER                          LOC           LOC



    Considered as a classification problem




                                                                     4
Related Work (1)
• Supervised Learning
    Many supervised machine learning approaches have been
     successfully applied
      • (Kambhatla, 2004; Zhou et al., 2005; Zelenko et al., 2003; Culotta and
        Sorensen, 2004; Bunescu and Mooney, 2005; Zhang et al., 2006)

• Semi-supervised Learning
    To obtain the annotations of unlabeled instances from the seed
     information
      • (Brin, 1999; Riloff and Jones, 1999; Agichtein and Gravano, 2000;
        Sudo et al, 2003; Yangarber, 2003; Stevenson and Greenwood, 2006;
        Zhang, 2004; Chen el al., 2006; Zhou et al., 2009)



                                                                             5
Motivation
• Resources for Relation Extraction
    Supervised/Semi-supervised Approaches
      • Labeled corpora for supervised learning
      • Seed instances for semi-supervised learning
      • Available for only a few languages
           ACE 2003 Multilingual Training Dataset
              • English (252 articles)
              • Chinese (221 articles)
              • Arabic (206 articles)
      • No resources for other languages
           Korean




                                                      6
Related Work (2)
• Self-supervised Learning
    To obtain the annotated dataset without any human effort
    Using the information obtained from external resources
       • Heuristic-based Method (Banko et al., 2007; Banko et al., 2008)
       • Wikipedia-based Methods (Wu and Weld, 2010)

• Cross-lingual Annotation Projection
    To leverage parallel corpora to project the relation annotations on
     the resource-rich source language to the resource-poor target
     language (Kim et al., 2010, Kim et al., 2011)




                                                                           7
Contents
• Introduction
• Methods
    Cross-lingual Annotation Projection for Relation Extraction
    Graph-based Projection Approach
• Implementation
• Evaluation
• Conclusions




                                                               8
Overall Architecture
Annotation                Parallel
                                                     Projection
                          Corpus


         Sentences in                 Sentences in
                 Ls                        Lt



        Preprocessing                Preprocessing
        (POS Tagging,                (POS Tagging,
           Parsing)                     Parsing)




               NER                   Word Alignment




              Relation
                                       Projection
             Extraction



          Annotated                    Annotated
         Sentences in                 Sentences in
                 Ls                        Lt                     9
Direct Projection
                                                                                (Kim et al., 2010)
• Annotation


• Projection




                        fE (<Barack Obama, Honolulu>) = 1
       Barack Obama              was born in        Honolulu          ,    Hawaii         .


    버락 오바마               는       하와이         의      호놀룰루              에서          태어났다
    (beo-rak-o-ba-ma)   (neun)   (ha-wa-i)   (ui)   (ho-nol-rul-ru)   (e-seo)     (tae-eo-nat-da)


                          fK (<버락 오바마, 호놀룰루>) = 1
                                                                                                    10
Limitations of Direct Projection
• Direct projection approach is still vulnerable to the
  erroneous inputs generated by preprocessors
• Main causes of this limitation
    Considering alignment between entity candidates only, not any
     contextual information
    Performed by just a single pass process




                                                                     11
Graph-based Learning
• Semi-supervised learning algorithm
• Defining a graph
    The nodes represent labeled and unlabeled examples in a dataset
    The edges reflect the similarity of examples
• Learning a labeling function in an iterative manner
    It should be close to the given labels on the similar labeled nodes
    It should be smooth on the whole graph
• Related Work
    Graph-based Learning for Relation Extraction (Chen et al, 2006)
    Bilingual projection of POS tagging (Das and Petrov, 2011)


                                                                           12
Graph Construction
• Graph Nodes
   Instance Nodes
      • Defined for all pairs of entity candidates in both languages
      • Each instance node has a soft label vector Y = [y+ y-]
   Context Nodes
      • For identifying the relation descriptors of the positive instances
      • Defined for each trigram which is located between a given entity pair
        which is semantically related
      • Each context node has a soft label vector Y = [y+ y-]


                       <ARG1> was born in <ARG2>



   <ARG1> was born                was born in                born in <ARG2>     13
Graph Construction
• Edge Weights
   Between instance node and context node in the same language
        𝑤 𝑣 𝑖,𝑗 , 𝑢 𝑘
                             1             𝑖𝑓 𝑣 𝑖𝑗 ℎ𝑎𝑠 𝑢 𝑘 𝑎𝑠 𝑎 𝑐𝑜𝑛𝑡𝑒𝑥𝑡𝑢𝑎𝑙 𝑠𝑢𝑏𝑠𝑒𝑞𝑢𝑒𝑛𝑐𝑒,
                           = 0                                𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒.
   Between context nodes in a language
             𝑘,
                                                   |𝑢 𝑘 ∩ 𝑢 𝑙 |
       𝑤(𝑢        𝑢 𝑙)     = 𝐽(𝑢      𝑘,
                                           𝑢 𝑙)   = 𝑘           .
                                                   |𝑢 ∪ 𝑢 𝑙 |

   Between context nodes in source and target languages
                                      𝑐𝑜𝑢𝑛𝑡 𝑢 𝑠𝑘 , 𝑢 𝑙𝑡
       𝑤(𝑢 𝑠𝑘 ,   𝑢 𝑙𝑡 )   =                      𝑘       𝑚
                                                               ,
                                 𝑢𝑡   𝑚 ‍ 𝑐𝑜𝑢𝑛𝑡 𝑢 𝑠 , 𝑢 𝑡




                                                                                          14
Graph Construction
• Example




                                 15
Label Propagation
                                       Initialize T
• Algorithm
    Input
      • A transition matrix T
      • An initial label matrix Y0    Normalize T
    Output
      • The updated label matrix Yt

                                       Initialize Y




                                        Update Y




                                                 16
Label Propagation
• Executed in three phases




               1st phase




                             2nd phase




                                         3rd phase




                                                     17
Contents
• Introduction
• Methods
    Cross-lingual Annotation Projection for Relation Extraction
    Graph-based Projection Approach
• Evaluation
• Conclusions




                                                                   18
Implementation
• Dataset
    English-Korean parallel corpus
       • 266,982 bi-sentence pairs in English and Korean
       • Aligned by GIZA++
• Annotation
    ReVerb (Fader et al., 2011)
       • English Open IE system
• Label Propagation
    Junto Label Propagation Toolkit
• Learning
    Tree kernel-based SVM classifier
       • Shortest path dependency kernel (Bunescu and Mooney, 2005)
       • SVM-Light (Joachims, 1998)

                                                                      19
Evaluation
• Dataset
    Manually annotated Korean dataset
      • Obtained from the Web following Bunescu and Mooney(2007)’s work
      • 500 sentences with manual annotations for four relation types
             Acquisition
             Birthplace
             Inventor Of
             Won Prize

• Evaluation Metrics
    Precision/Recall/F-measure




                                                                          20
Experimental Results
• Direct Projection vs. Graph-based Projection


                   Direct Projection     Graph-based Projection
      Type
                  P       R        F       P       R       F
   Acquisition   51.6    87.7     64.9    55.3    91.2    68.9
   Birthplace    69.8    84.5     76.4    73.8    87.3    80.0
   Inventor of   62.4    85.3     72.1    66.3    89.7    76.3
   Won Prize     73.3    80.5     76.7    76.4    82.9    79.5
      Total      63.9    84.2     72.7    67.7    87.4    76.3




                                                                 21
Experimental Results
• Comparisons to other self-supervised approaches
    Heuristic-based Approach (Banko et al., 2007; Banko et al., 2008)
       • Korean Treebank and Syntactic Heuristics
    Wikipedia-based Approach (Wu and Weld, 2010)
       • Korean Wikipedia articles and Infoboxes


          Approach                    P              R        F

       Heuristic-based              92.31           17.27   29.09

      Wikipedia-based               66.67           66.91   66.79

      Projection-based             67.69            87.41   76.30

                                                                         22
Contents
• Introduction
• Methods
    Cross-lingual Annotation Projection for Relation Extraction
    Graph-based Projection Approach
• Evaluation
• Conclusions




                                                                   23
Conclusion
• Summary
    A graph-based projection approach for relation extraction
       • Label propagation algorithm
       • On a graph that represents the instance and context features of both
         the source and target languages
    Experimental results show that our approach helps to improve the
     performances of relation extraction compared to other approaches
• Future work
    To relieve the high complexity problem of the approach
    To deal with more expanded graph structure to improve the
     extraction performances


                                                                                24
Q&A
1 of 25

Recommended

A Cross-Lingual Annotation Projection Approach for Relation Detection by
A Cross-Lingual Annotation Projection Approach for Relation DetectionA Cross-Lingual Annotation Projection Approach for Relation Detection
A Cross-Lingual Annotation Projection Approach for Relation DetectionSeokhwan Kim
602 views33 slides
BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding by
BERT: Pre-training of Deep Bidirectional Transformers for Language UnderstandingBERT: Pre-training of Deep Bidirectional Transformers for Language Understanding
BERT: Pre-training of Deep Bidirectional Transformers for Language UnderstandingMinh Quang-Nhat Pham
8.2K views40 slides
Bert pre_training_of_deep_bidirectional_transformers_for_language_understanding by
Bert  pre_training_of_deep_bidirectional_transformers_for_language_understandingBert  pre_training_of_deep_bidirectional_transformers_for_language_understanding
Bert pre_training_of_deep_bidirectional_transformers_for_language_understandingThyrixYang1
224 views21 slides
1909 paclic by
1909 paclic1909 paclic
1909 paclicWarNik Chow
311 views24 slides
Pre trained language model by
Pre trained language modelPre trained language model
Pre trained language modelJiWenKim
365 views25 slides
2010 PACLIC - pay attention to categories by
2010 PACLIC - pay attention to categories2010 PACLIC - pay attention to categories
2010 PACLIC - pay attention to categoriesWarNik Chow
62 views21 slides

More Related Content

What's hot

Ontology Integration and Interoperability (OntoIOp) – Part 1: The Distributed... by
Ontology Integration and Interoperability (OntoIOp) – Part 1: The Distributed...Ontology Integration and Interoperability (OntoIOp) – Part 1: The Distributed...
Ontology Integration and Interoperability (OntoIOp) – Part 1: The Distributed...Christoph Lange
790 views14 slides
NLP from scratch by
NLP from scratch NLP from scratch
NLP from scratch Bryan Gummibearehausen
1.5K views56 slides
A System for the Recognition of Handwritten Yorùbá Characters by
A System for the Recognition of Handwritten Yorùbá CharactersA System for the Recognition of Handwritten Yorùbá Characters
A System for the Recognition of Handwritten Yorùbá CharactersGuy De Pauw
1.3K views45 slides
Deep Learning for NLP: An Introduction to Neural Word Embeddings by
Deep Learning for NLP: An Introduction to Neural Word EmbeddingsDeep Learning for NLP: An Introduction to Neural Word Embeddings
Deep Learning for NLP: An Introduction to Neural Word EmbeddingsRoelof Pieters
20.1K views75 slides
NLP State of the Art | BERT by
NLP State of the Art | BERTNLP State of the Art | BERT
NLP State of the Art | BERTshaurya uppal
2.5K views26 slides
BERT: Bidirectional Encoder Representations from Transformers by
BERT: Bidirectional Encoder Representations from TransformersBERT: Bidirectional Encoder Representations from Transformers
BERT: Bidirectional Encoder Representations from TransformersLiangqun Lu
1.9K views46 slides

What's hot(20)

Ontology Integration and Interoperability (OntoIOp) – Part 1: The Distributed... by Christoph Lange
Ontology Integration and Interoperability (OntoIOp) – Part 1: The Distributed...Ontology Integration and Interoperability (OntoIOp) – Part 1: The Distributed...
Ontology Integration and Interoperability (OntoIOp) – Part 1: The Distributed...
Christoph Lange790 views
A System for the Recognition of Handwritten Yorùbá Characters by Guy De Pauw
A System for the Recognition of Handwritten Yorùbá CharactersA System for the Recognition of Handwritten Yorùbá Characters
A System for the Recognition of Handwritten Yorùbá Characters
Guy De Pauw1.3K views
Deep Learning for NLP: An Introduction to Neural Word Embeddings by Roelof Pieters
Deep Learning for NLP: An Introduction to Neural Word EmbeddingsDeep Learning for NLP: An Introduction to Neural Word Embeddings
Deep Learning for NLP: An Introduction to Neural Word Embeddings
Roelof Pieters20.1K views
NLP State of the Art | BERT by shaurya uppal
NLP State of the Art | BERTNLP State of the Art | BERT
NLP State of the Art | BERT
shaurya uppal2.5K views
BERT: Bidirectional Encoder Representations from Transformers by Liangqun Lu
BERT: Bidirectional Encoder Representations from TransformersBERT: Bidirectional Encoder Representations from Transformers
BERT: Bidirectional Encoder Representations from Transformers
Liangqun Lu1.9K views
S-CUBE LP: Executing the HOCL: Concept of a Chemical Interpreter by virtual-campus
S-CUBE LP: Executing the HOCL: Concept of a Chemical InterpreterS-CUBE LP: Executing the HOCL: Concept of a Chemical Interpreter
S-CUBE LP: Executing the HOCL: Concept of a Chemical Interpreter
virtual-campus436 views
Ontology Integration and Interoperability (OntoIOp) – Part 1: The Distributed... by Christoph Lange
Ontology Integration and Interoperability (OntoIOp) – Part 1: The Distributed...Ontology Integration and Interoperability (OntoIOp) – Part 1: The Distributed...
Ontology Integration and Interoperability (OntoIOp) – Part 1: The Distributed...
Christoph Lange583 views
Grammarly AI-NLP Club #6 - Sequence Tagging using Neural Networks - Artem Che... by Grammarly
Grammarly AI-NLP Club #6 - Sequence Tagging using Neural Networks - Artem Che...Grammarly AI-NLP Club #6 - Sequence Tagging using Neural Networks - Artem Che...
Grammarly AI-NLP Club #6 - Sequence Tagging using Neural Networks - Artem Che...
Grammarly3.9K views
Multitier holistic Approach for urdu Nastaliq Recognition by Dr. Syed Hassan Amin
Multitier holistic Approach for urdu Nastaliq RecognitionMultitier holistic Approach for urdu Nastaliq Recognition
Multitier holistic Approach for urdu Nastaliq Recognition
Deep Reinforcement Learning with Distributional Semantic Rewards for Abstract... by Deren Lei
Deep Reinforcement Learning with Distributional Semantic Rewards for Abstract...Deep Reinforcement Learning with Distributional Semantic Rewards for Abstract...
Deep Reinforcement Learning with Distributional Semantic Rewards for Abstract...
Deren Lei148 views
A NOVEL APPROACH FOR NAMED ENTITY RECOGNITION ON HINDI LANGUAGE USING RESIDUA... by kevig
A NOVEL APPROACH FOR NAMED ENTITY RECOGNITION ON HINDI LANGUAGE USING RESIDUA...A NOVEL APPROACH FOR NAMED ENTITY RECOGNITION ON HINDI LANGUAGE USING RESIDUA...
A NOVEL APPROACH FOR NAMED ENTITY RECOGNITION ON HINDI LANGUAGE USING RESIDUA...
kevig67 views
VOC real world enterprise needs by Ivan Berlocher
VOC real world enterprise needsVOC real world enterprise needs
VOC real world enterprise needs
Ivan Berlocher514 views
Pattern Mining To Unknown Word Extraction (10 by Jason Yang
Pattern Mining To Unknown Word Extraction (10Pattern Mining To Unknown Word Extraction (10
Pattern Mining To Unknown Word Extraction (10
Jason Yang456 views
Representation Learning of Vectors of Words and Phrases by Felipe Moraes
Representation Learning of Vectors of Words and PhrasesRepresentation Learning of Vectors of Words and Phrases
Representation Learning of Vectors of Words and Phrases
Felipe Moraes6.4K views
BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding by gohyunwoong
BERT: Pre-training of Deep Bidirectional Transformers for Language UnderstandingBERT: Pre-training of Deep Bidirectional Transformers for Language Understanding
BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding
gohyunwoong860 views

Similar to A Graph-based Cross-lingual Projection Approach for Weakly Supervised Relation Extraction

A Cross-lingual Annotation Projection-based Self-supervision Approach for Ope... by
A Cross-lingual Annotation Projection-based Self-supervision Approach for Ope...A Cross-lingual Annotation Projection-based Self-supervision Approach for Ope...
A Cross-lingual Annotation Projection-based Self-supervision Approach for Ope...Seokhwan Kim
471 views41 slides
Semi-Supervised Autoencoders for Predicting Sentiment Distributions(第 5 回 De... by
 Semi-Supervised Autoencoders for Predicting Sentiment Distributions(第 5 回 De... Semi-Supervised Autoencoders for Predicting Sentiment Distributions(第 5 回 De...
Semi-Supervised Autoencoders for Predicting Sentiment Distributions(第 5 回 De...Ohsawa Goodfellow
17.3K views16 slides
Word Segmentation and Lexical Normalization for Unsegmented Languages by
Word Segmentation and Lexical Normalization for Unsegmented LanguagesWord Segmentation and Lexical Normalization for Unsegmented Languages
Word Segmentation and Lexical Normalization for Unsegmented Languageshs0041
726 views57 slides
Deep Learning and Modern Natural Language Processing (AnacondaCon2019) by
Deep Learning and Modern Natural Language Processing (AnacondaCon2019)Deep Learning and Modern Natural Language Processing (AnacondaCon2019)
Deep Learning and Modern Natural Language Processing (AnacondaCon2019)Zachary S. Brown
209 views53 slides
Erlang/OTP for Rubyists by
Erlang/OTP for RubyistsErlang/OTP for Rubyists
Erlang/OTP for RubyistsSean Cribbs
1.6K views51 slides
Engineering Intelligent NLP Applications Using Deep Learning – Part 2 by
Engineering Intelligent NLP Applications Using Deep Learning – Part 2 Engineering Intelligent NLP Applications Using Deep Learning – Part 2
Engineering Intelligent NLP Applications Using Deep Learning – Part 2 Saurabh Kaushik
2.5K views30 slides

Similar to A Graph-based Cross-lingual Projection Approach for Weakly Supervised Relation Extraction(20)

A Cross-lingual Annotation Projection-based Self-supervision Approach for Ope... by Seokhwan Kim
A Cross-lingual Annotation Projection-based Self-supervision Approach for Ope...A Cross-lingual Annotation Projection-based Self-supervision Approach for Ope...
A Cross-lingual Annotation Projection-based Self-supervision Approach for Ope...
Seokhwan Kim471 views
Semi-Supervised Autoencoders for Predicting Sentiment Distributions(第 5 回 De... by Ohsawa Goodfellow
 Semi-Supervised Autoencoders for Predicting Sentiment Distributions(第 5 回 De... Semi-Supervised Autoencoders for Predicting Sentiment Distributions(第 5 回 De...
Semi-Supervised Autoencoders for Predicting Sentiment Distributions(第 5 回 De...
Ohsawa Goodfellow17.3K views
Word Segmentation and Lexical Normalization for Unsegmented Languages by hs0041
Word Segmentation and Lexical Normalization for Unsegmented LanguagesWord Segmentation and Lexical Normalization for Unsegmented Languages
Word Segmentation and Lexical Normalization for Unsegmented Languages
hs0041726 views
Deep Learning and Modern Natural Language Processing (AnacondaCon2019) by Zachary S. Brown
Deep Learning and Modern Natural Language Processing (AnacondaCon2019)Deep Learning and Modern Natural Language Processing (AnacondaCon2019)
Deep Learning and Modern Natural Language Processing (AnacondaCon2019)
Zachary S. Brown209 views
Erlang/OTP for Rubyists by Sean Cribbs
Erlang/OTP for RubyistsErlang/OTP for Rubyists
Erlang/OTP for Rubyists
Sean Cribbs1.6K views
Engineering Intelligent NLP Applications Using Deep Learning – Part 2 by Saurabh Kaushik
Engineering Intelligent NLP Applications Using Deep Learning – Part 2 Engineering Intelligent NLP Applications Using Deep Learning – Part 2
Engineering Intelligent NLP Applications Using Deep Learning – Part 2
Saurabh Kaushik2.5K views
Deep Learning勉強会@小町研 "Learning Character-level Representations for Part-of-Sp... by Yuki Tomo
Deep Learning勉強会@小町研 "Learning Character-level Representations for Part-of-Sp...Deep Learning勉強会@小町研 "Learning Character-level Representations for Part-of-Sp...
Deep Learning勉強会@小町研 "Learning Character-level Representations for Part-of-Sp...
Yuki Tomo1.4K views
Nltk natural language toolkit overview and application @ PyCon.tw 2012 by Jimmy Lai
Nltk  natural language toolkit overview and application @ PyCon.tw 2012Nltk  natural language toolkit overview and application @ PyCon.tw 2012
Nltk natural language toolkit overview and application @ PyCon.tw 2012
Jimmy Lai2.9K views
TSD2013 PPT.AUTOMATIC MACHINE TRANSLATION EVALUATION WITH PART-OF-SPEECH INFO... by Lifeng (Aaron) Han
TSD2013 PPT.AUTOMATIC MACHINE TRANSLATION EVALUATION WITH PART-OF-SPEECH INFO...TSD2013 PPT.AUTOMATIC MACHINE TRANSLATION EVALUATION WITH PART-OF-SPEECH INFO...
TSD2013 PPT.AUTOMATIC MACHINE TRANSLATION EVALUATION WITH PART-OF-SPEECH INFO...
Lifeng (Aaron) Han952 views
Colloquium talk on modal sense classification using a convolutional neural ne... by Ana Marasović
Colloquium talk on modal sense classification using a convolutional neural ne...Colloquium talk on modal sense classification using a convolutional neural ne...
Colloquium talk on modal sense classification using a convolutional neural ne...
Ana Marasović72 views
Incremental Difference as Feature for Lipreading by IDES Editor
Incremental Difference as Feature for LipreadingIncremental Difference as Feature for Lipreading
Incremental Difference as Feature for Lipreading
IDES Editor391 views
ESWC-2011: S-Match received 7 years award by Pavel Shvaiko
ESWC-2011: S-Match received 7 years award ESWC-2011: S-Match received 7 years award
ESWC-2011: S-Match received 7 years award
Pavel Shvaiko547 views
CUHK intern PPT. Machine Translation Evaluation: Methods and Tools by Lifeng (Aaron) Han
CUHK intern PPT. Machine Translation Evaluation: Methods and Tools CUHK intern PPT. Machine Translation Evaluation: Methods and Tools
CUHK intern PPT. Machine Translation Evaluation: Methods and Tools
Lifeng (Aaron) Han1.7K views

More from Seokhwan Kim

The Eighth Dialog System Technology Challenge (DSTC8) by
The Eighth Dialog System Technology Challenge (DSTC8)The Eighth Dialog System Technology Challenge (DSTC8)
The Eighth Dialog System Technology Challenge (DSTC8)Seokhwan Kim
337 views1 slide
Deep Recurrent Neural Networks with Layer-wise Multi-head Attentions for Punc... by
Deep Recurrent Neural Networks with Layer-wise Multi-head Attentions for Punc...Deep Recurrent Neural Networks with Layer-wise Multi-head Attentions for Punc...
Deep Recurrent Neural Networks with Layer-wise Multi-head Attentions for Punc...Seokhwan Kim
285 views1 slide
Dynamic Memory Networks for Dialogue Topic Tracking by
Dynamic Memory Networks for Dialogue Topic TrackingDynamic Memory Networks for Dialogue Topic Tracking
Dynamic Memory Networks for Dialogue Topic TrackingSeokhwan Kim
421 views1 slide
The Fifth Dialog State Tracking Challenge (DSTC5) by
The Fifth Dialog State Tracking Challenge (DSTC5)The Fifth Dialog State Tracking Challenge (DSTC5)
The Fifth Dialog State Tracking Challenge (DSTC5)Seokhwan Kim
645 views1 slide
Natural Language in Human-Robot Interaction by
Natural Language in Human-Robot InteractionNatural Language in Human-Robot Interaction
Natural Language in Human-Robot InteractionSeokhwan Kim
4.2K views169 slides
Exploring Convolutional and Recurrent Neural Networks in Sequential Labelling... by
Exploring Convolutional and Recurrent Neural Networks in Sequential Labelling...Exploring Convolutional and Recurrent Neural Networks in Sequential Labelling...
Exploring Convolutional and Recurrent Neural Networks in Sequential Labelling...Seokhwan Kim
475 views1 slide

More from Seokhwan Kim(19)

The Eighth Dialog System Technology Challenge (DSTC8) by Seokhwan Kim
The Eighth Dialog System Technology Challenge (DSTC8)The Eighth Dialog System Technology Challenge (DSTC8)
The Eighth Dialog System Technology Challenge (DSTC8)
Seokhwan Kim337 views
Deep Recurrent Neural Networks with Layer-wise Multi-head Attentions for Punc... by Seokhwan Kim
Deep Recurrent Neural Networks with Layer-wise Multi-head Attentions for Punc...Deep Recurrent Neural Networks with Layer-wise Multi-head Attentions for Punc...
Deep Recurrent Neural Networks with Layer-wise Multi-head Attentions for Punc...
Seokhwan Kim285 views
Dynamic Memory Networks for Dialogue Topic Tracking by Seokhwan Kim
Dynamic Memory Networks for Dialogue Topic TrackingDynamic Memory Networks for Dialogue Topic Tracking
Dynamic Memory Networks for Dialogue Topic Tracking
Seokhwan Kim421 views
The Fifth Dialog State Tracking Challenge (DSTC5) by Seokhwan Kim
The Fifth Dialog State Tracking Challenge (DSTC5)The Fifth Dialog State Tracking Challenge (DSTC5)
The Fifth Dialog State Tracking Challenge (DSTC5)
Seokhwan Kim645 views
Natural Language in Human-Robot Interaction by Seokhwan Kim
Natural Language in Human-Robot InteractionNatural Language in Human-Robot Interaction
Natural Language in Human-Robot Interaction
Seokhwan Kim4.2K views
Exploring Convolutional and Recurrent Neural Networks in Sequential Labelling... by Seokhwan Kim
Exploring Convolutional and Recurrent Neural Networks in Sequential Labelling...Exploring Convolutional and Recurrent Neural Networks in Sequential Labelling...
Exploring Convolutional and Recurrent Neural Networks in Sequential Labelling...
Seokhwan Kim475 views
The Fourth Dialog State Tracking Challenge (DSTC4) by Seokhwan Kim
The Fourth Dialog State Tracking Challenge (DSTC4)The Fourth Dialog State Tracking Challenge (DSTC4)
The Fourth Dialog State Tracking Challenge (DSTC4)
Seokhwan Kim1K views
Wikification of Concept Mentions within Spoken Dialogues Using Domain Constra... by Seokhwan Kim
Wikification of Concept Mentions within Spoken Dialogues Using Domain Constra...Wikification of Concept Mentions within Spoken Dialogues Using Domain Constra...
Wikification of Concept Mentions within Spoken Dialogues Using Domain Constra...
Seokhwan Kim476 views
Towards Improving Dialogue Topic Tracking Performances with Wikification of C... by Seokhwan Kim
Towards Improving Dialogue Topic Tracking Performances with Wikification of C...Towards Improving Dialogue Topic Tracking Performances with Wikification of C...
Towards Improving Dialogue Topic Tracking Performances with Wikification of C...
Seokhwan Kim871 views
A Composite Kernel Approach for Dialog Topic Tracking with Structured Domain ... by Seokhwan Kim
A Composite Kernel Approach for Dialog Topic Tracking with Structured Domain ...A Composite Kernel Approach for Dialog Topic Tracking with Structured Domain ...
A Composite Kernel Approach for Dialog Topic Tracking with Structured Domain ...
Seokhwan Kim610 views
Sequential Labeling for Tracking Dynamic Dialog States by Seokhwan Kim
Sequential Labeling for Tracking Dynamic Dialog StatesSequential Labeling for Tracking Dynamic Dialog States
Sequential Labeling for Tracking Dynamic Dialog States
Seokhwan Kim525 views
Wikipedia-based Kernels for Dialogue Topic Tracking by Seokhwan Kim
Wikipedia-based Kernels for Dialogue Topic TrackingWikipedia-based Kernels for Dialogue Topic Tracking
Wikipedia-based Kernels for Dialogue Topic Tracking
Seokhwan Kim975 views
A Graph-based Cross-lingual Projection Approach for Spoken Language Understan... by Seokhwan Kim
A Graph-based Cross-lingual Projection Approach for Spoken Language Understan...A Graph-based Cross-lingual Projection Approach for Spoken Language Understan...
A Graph-based Cross-lingual Projection Approach for Spoken Language Understan...
Seokhwan Kim918 views
MMR-based active machine learning for Bio named entity recognition by Seokhwan Kim
MMR-based active machine learning for Bio named entity recognitionMMR-based active machine learning for Bio named entity recognition
MMR-based active machine learning for Bio named entity recognition
Seokhwan Kim489 views
A semi-supervised method for efficient construction of statistical spoken lan... by Seokhwan Kim
A semi-supervised method for efficient construction of statistical spoken lan...A semi-supervised method for efficient construction of statistical spoken lan...
A semi-supervised method for efficient construction of statistical spoken lan...
Seokhwan Kim368 views
A spoken dialog system for electronic program guide information access by Seokhwan Kim
A spoken dialog system for electronic program guide information accessA spoken dialog system for electronic program guide information access
A spoken dialog system for electronic program guide information access
Seokhwan Kim487 views
An alignment-based approach to semi-supervised relation extraction including ... by Seokhwan Kim
An alignment-based approach to semi-supervised relation extraction including ...An alignment-based approach to semi-supervised relation extraction including ...
An alignment-based approach to semi-supervised relation extraction including ...
Seokhwan Kim414 views
An Alignment-based Pattern Representation Model for Information Extraction by Seokhwan Kim
An Alignment-based Pattern Representation Model for Information ExtractionAn Alignment-based Pattern Representation Model for Information Extraction
An Alignment-based Pattern Representation Model for Information Extraction
Seokhwan Kim413 views
EPG 정보 검색을 위한 예제 기반 자연어 대화 시스템 by Seokhwan Kim
EPG 정보 검색을 위한 예제 기반 자연어 대화 시스템EPG 정보 검색을 위한 예제 기반 자연어 대화 시스템
EPG 정보 검색을 위한 예제 기반 자연어 대화 시스템
Seokhwan Kim1.2K views

Recently uploaded

Digital Product-Centric Enterprise and Enterprise Architecture - Tan Eng Tsze by
Digital Product-Centric Enterprise and Enterprise Architecture - Tan Eng TszeDigital Product-Centric Enterprise and Enterprise Architecture - Tan Eng Tsze
Digital Product-Centric Enterprise and Enterprise Architecture - Tan Eng TszeNUS-ISS
19 views47 slides
How to reduce cold starts for Java Serverless applications in AWS at JCON Wor... by
How to reduce cold starts for Java Serverless applications in AWS at JCON Wor...How to reduce cold starts for Java Serverless applications in AWS at JCON Wor...
How to reduce cold starts for Java Serverless applications in AWS at JCON Wor...Vadym Kazulkin
75 views64 slides
Attacking IoT Devices from a Web Perspective - Linux Day by
Attacking IoT Devices from a Web Perspective - Linux Day Attacking IoT Devices from a Web Perspective - Linux Day
Attacking IoT Devices from a Web Perspective - Linux Day Simone Onofri
15 views68 slides
Business Analyst Series 2023 - Week 3 Session 5 by
Business Analyst Series 2023 -  Week 3 Session 5Business Analyst Series 2023 -  Week 3 Session 5
Business Analyst Series 2023 - Week 3 Session 5DianaGray10
209 views20 slides
Architecting CX Measurement Frameworks and Ensuring CX Metrics are fit for Pu... by
Architecting CX Measurement Frameworks and Ensuring CX Metrics are fit for Pu...Architecting CX Measurement Frameworks and Ensuring CX Metrics are fit for Pu...
Architecting CX Measurement Frameworks and Ensuring CX Metrics are fit for Pu...NUS-ISS
37 views54 slides
Understanding GenAI/LLM and What is Google Offering - Felix Goh by
Understanding GenAI/LLM and What is Google Offering - Felix GohUnderstanding GenAI/LLM and What is Google Offering - Felix Goh
Understanding GenAI/LLM and What is Google Offering - Felix GohNUS-ISS
41 views33 slides

Recently uploaded(20)

Digital Product-Centric Enterprise and Enterprise Architecture - Tan Eng Tsze by NUS-ISS
Digital Product-Centric Enterprise and Enterprise Architecture - Tan Eng TszeDigital Product-Centric Enterprise and Enterprise Architecture - Tan Eng Tsze
Digital Product-Centric Enterprise and Enterprise Architecture - Tan Eng Tsze
NUS-ISS19 views
How to reduce cold starts for Java Serverless applications in AWS at JCON Wor... by Vadym Kazulkin
How to reduce cold starts for Java Serverless applications in AWS at JCON Wor...How to reduce cold starts for Java Serverless applications in AWS at JCON Wor...
How to reduce cold starts for Java Serverless applications in AWS at JCON Wor...
Vadym Kazulkin75 views
Attacking IoT Devices from a Web Perspective - Linux Day by Simone Onofri
Attacking IoT Devices from a Web Perspective - Linux Day Attacking IoT Devices from a Web Perspective - Linux Day
Attacking IoT Devices from a Web Perspective - Linux Day
Simone Onofri15 views
Business Analyst Series 2023 - Week 3 Session 5 by DianaGray10
Business Analyst Series 2023 -  Week 3 Session 5Business Analyst Series 2023 -  Week 3 Session 5
Business Analyst Series 2023 - Week 3 Session 5
DianaGray10209 views
Architecting CX Measurement Frameworks and Ensuring CX Metrics are fit for Pu... by NUS-ISS
Architecting CX Measurement Frameworks and Ensuring CX Metrics are fit for Pu...Architecting CX Measurement Frameworks and Ensuring CX Metrics are fit for Pu...
Architecting CX Measurement Frameworks and Ensuring CX Metrics are fit for Pu...
NUS-ISS37 views
Understanding GenAI/LLM and What is Google Offering - Felix Goh by NUS-ISS
Understanding GenAI/LLM and What is Google Offering - Felix GohUnderstanding GenAI/LLM and What is Google Offering - Felix Goh
Understanding GenAI/LLM and What is Google Offering - Felix Goh
NUS-ISS41 views
Spesifikasi Lengkap ASUS Vivobook Go 14 by Dot Semarang
Spesifikasi Lengkap ASUS Vivobook Go 14Spesifikasi Lengkap ASUS Vivobook Go 14
Spesifikasi Lengkap ASUS Vivobook Go 14
Dot Semarang35 views
The details of description: Techniques, tips, and tangents on alternative tex... by BookNet Canada
The details of description: Techniques, tips, and tangents on alternative tex...The details of description: Techniques, tips, and tangents on alternative tex...
The details of description: Techniques, tips, and tangents on alternative tex...
BookNet Canada121 views
PharoJS - Zürich Smalltalk Group Meetup November 2023 by Noury Bouraqadi
PharoJS - Zürich Smalltalk Group Meetup November 2023PharoJS - Zürich Smalltalk Group Meetup November 2023
PharoJS - Zürich Smalltalk Group Meetup November 2023
Noury Bouraqadi120 views
Combining Orchestration and Choreography for a Clean Architecture by ThomasHeinrichs1
Combining Orchestration and Choreography for a Clean ArchitectureCombining Orchestration and Choreography for a Clean Architecture
Combining Orchestration and Choreography for a Clean Architecture
ThomasHeinrichs169 views
handbook for web 3 adoption.pdf by Liveplex
handbook for web 3 adoption.pdfhandbook for web 3 adoption.pdf
handbook for web 3 adoption.pdf
Liveplex19 views
[2023] Putting the R! in R&D.pdf by Eleanor McHugh
[2023] Putting the R! in R&D.pdf[2023] Putting the R! in R&D.pdf
[2023] Putting the R! in R&D.pdf
Eleanor McHugh38 views
STPI OctaNE CoE Brochure.pdf by madhurjyapb
STPI OctaNE CoE Brochure.pdfSTPI OctaNE CoE Brochure.pdf
STPI OctaNE CoE Brochure.pdf
madhurjyapb12 views
Empathic Computing: Delivering the Potential of the Metaverse by Mark Billinghurst
Empathic Computing: Delivering  the Potential of the MetaverseEmpathic Computing: Delivering  the Potential of the Metaverse
Empathic Computing: Delivering the Potential of the Metaverse
Mark Billinghurst470 views
.conf Go 2023 - Data analysis as a routine by Splunk
.conf Go 2023 - Data analysis as a routine.conf Go 2023 - Data analysis as a routine
.conf Go 2023 - Data analysis as a routine
Splunk93 views
DALI Basics Course 2023 by Ivory Egg
DALI Basics Course  2023DALI Basics Course  2023
DALI Basics Course 2023
Ivory Egg14 views
Beyond the Hype: What Generative AI Means for the Future of Work - Damien Cum... by NUS-ISS
Beyond the Hype: What Generative AI Means for the Future of Work - Damien Cum...Beyond the Hype: What Generative AI Means for the Future of Work - Damien Cum...
Beyond the Hype: What Generative AI Means for the Future of Work - Damien Cum...
NUS-ISS34 views

A Graph-based Cross-lingual Projection Approach for Weakly Supervised Relation Extraction

  • 1. A GRAPH-BASED CROSS-LINGUAL PROJECTION APPROACH FOR WEAKLY SUPERVISED RELATION EXTRACTION The 50th Annual Meeting of the Association for Computational Linguistics (ACL 2012) July 11th, 2012, Jeju Seokhwan Kim (Institute for Infocomm Research) Gary Geunbae Lee (POSTECH)
  • 2. Contents • Introduction • Methods  Cross-lingual Annotation Projection for Relation Extraction  Graph-based Projection Approach • Evaluation • Conclusions 2
  • 3. Contents • Introduction • Methods  Cross-lingual Annotation Projection for Relation Extraction  Graph-based Projection Approach • Evaluation • Conclusions 3
  • 4. Problem Definition • Relation Extraction  To identify semantic relations between a pair of entities Birthplace Barack Obama was born in Honolulu , Hawaii . PER LOC LOC  Considered as a classification problem 4
  • 5. Related Work (1) • Supervised Learning  Many supervised machine learning approaches have been successfully applied • (Kambhatla, 2004; Zhou et al., 2005; Zelenko et al., 2003; Culotta and Sorensen, 2004; Bunescu and Mooney, 2005; Zhang et al., 2006) • Semi-supervised Learning  To obtain the annotations of unlabeled instances from the seed information • (Brin, 1999; Riloff and Jones, 1999; Agichtein and Gravano, 2000; Sudo et al, 2003; Yangarber, 2003; Stevenson and Greenwood, 2006; Zhang, 2004; Chen el al., 2006; Zhou et al., 2009) 5
  • 6. Motivation • Resources for Relation Extraction  Supervised/Semi-supervised Approaches • Labeled corpora for supervised learning • Seed instances for semi-supervised learning • Available for only a few languages  ACE 2003 Multilingual Training Dataset • English (252 articles) • Chinese (221 articles) • Arabic (206 articles) • No resources for other languages  Korean 6
  • 7. Related Work (2) • Self-supervised Learning  To obtain the annotated dataset without any human effort  Using the information obtained from external resources • Heuristic-based Method (Banko et al., 2007; Banko et al., 2008) • Wikipedia-based Methods (Wu and Weld, 2010) • Cross-lingual Annotation Projection  To leverage parallel corpora to project the relation annotations on the resource-rich source language to the resource-poor target language (Kim et al., 2010, Kim et al., 2011) 7
  • 8. Contents • Introduction • Methods  Cross-lingual Annotation Projection for Relation Extraction  Graph-based Projection Approach • Implementation • Evaluation • Conclusions 8
  • 9. Overall Architecture Annotation Parallel Projection Corpus Sentences in Sentences in Ls Lt Preprocessing Preprocessing (POS Tagging, (POS Tagging, Parsing) Parsing) NER Word Alignment Relation Projection Extraction Annotated Annotated Sentences in Sentences in Ls Lt 9
  • 10. Direct Projection (Kim et al., 2010) • Annotation • Projection fE (<Barack Obama, Honolulu>) = 1 Barack Obama was born in Honolulu , Hawaii . 버락 오바마 는 하와이 의 호놀룰루 에서 태어났다 (beo-rak-o-ba-ma) (neun) (ha-wa-i) (ui) (ho-nol-rul-ru) (e-seo) (tae-eo-nat-da) fK (<버락 오바마, 호놀룰루>) = 1 10
  • 11. Limitations of Direct Projection • Direct projection approach is still vulnerable to the erroneous inputs generated by preprocessors • Main causes of this limitation  Considering alignment between entity candidates only, not any contextual information  Performed by just a single pass process 11
  • 12. Graph-based Learning • Semi-supervised learning algorithm • Defining a graph  The nodes represent labeled and unlabeled examples in a dataset  The edges reflect the similarity of examples • Learning a labeling function in an iterative manner  It should be close to the given labels on the similar labeled nodes  It should be smooth on the whole graph • Related Work  Graph-based Learning for Relation Extraction (Chen et al, 2006)  Bilingual projection of POS tagging (Das and Petrov, 2011) 12
  • 13. Graph Construction • Graph Nodes  Instance Nodes • Defined for all pairs of entity candidates in both languages • Each instance node has a soft label vector Y = [y+ y-]  Context Nodes • For identifying the relation descriptors of the positive instances • Defined for each trigram which is located between a given entity pair which is semantically related • Each context node has a soft label vector Y = [y+ y-] <ARG1> was born in <ARG2> <ARG1> was born was born in born in <ARG2> 13
  • 14. Graph Construction • Edge Weights  Between instance node and context node in the same language 𝑤 𝑣 𝑖,𝑗 , 𝑢 𝑘 1 𝑖𝑓 𝑣 𝑖𝑗 ℎ𝑎𝑠 𝑢 𝑘 𝑎𝑠 𝑎 𝑐𝑜𝑛𝑡𝑒𝑥𝑡𝑢𝑎𝑙 𝑠𝑢𝑏𝑠𝑒𝑞𝑢𝑒𝑛𝑐𝑒, = 0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒.  Between context nodes in a language 𝑘, |𝑢 𝑘 ∩ 𝑢 𝑙 | 𝑤(𝑢 𝑢 𝑙) = 𝐽(𝑢 𝑘, 𝑢 𝑙) = 𝑘 . |𝑢 ∪ 𝑢 𝑙 |  Between context nodes in source and target languages 𝑐𝑜𝑢𝑛𝑡 𝑢 𝑠𝑘 , 𝑢 𝑙𝑡 𝑤(𝑢 𝑠𝑘 , 𝑢 𝑙𝑡 ) = 𝑘 𝑚 , 𝑢𝑡 𝑚 ‍ 𝑐𝑜𝑢𝑛𝑡 𝑢 𝑠 , 𝑢 𝑡 14
  • 16. Label Propagation Initialize T • Algorithm  Input • A transition matrix T • An initial label matrix Y0 Normalize T  Output • The updated label matrix Yt Initialize Y Update Y 16
  • 17. Label Propagation • Executed in three phases 1st phase 2nd phase 3rd phase 17
  • 18. Contents • Introduction • Methods  Cross-lingual Annotation Projection for Relation Extraction  Graph-based Projection Approach • Evaluation • Conclusions 18
  • 19. Implementation • Dataset  English-Korean parallel corpus • 266,982 bi-sentence pairs in English and Korean • Aligned by GIZA++ • Annotation  ReVerb (Fader et al., 2011) • English Open IE system • Label Propagation  Junto Label Propagation Toolkit • Learning  Tree kernel-based SVM classifier • Shortest path dependency kernel (Bunescu and Mooney, 2005) • SVM-Light (Joachims, 1998) 19
  • 20. Evaluation • Dataset  Manually annotated Korean dataset • Obtained from the Web following Bunescu and Mooney(2007)’s work • 500 sentences with manual annotations for four relation types  Acquisition  Birthplace  Inventor Of  Won Prize • Evaluation Metrics  Precision/Recall/F-measure 20
  • 21. Experimental Results • Direct Projection vs. Graph-based Projection Direct Projection Graph-based Projection Type P R F P R F Acquisition 51.6 87.7 64.9 55.3 91.2 68.9 Birthplace 69.8 84.5 76.4 73.8 87.3 80.0 Inventor of 62.4 85.3 72.1 66.3 89.7 76.3 Won Prize 73.3 80.5 76.7 76.4 82.9 79.5 Total 63.9 84.2 72.7 67.7 87.4 76.3 21
  • 22. Experimental Results • Comparisons to other self-supervised approaches  Heuristic-based Approach (Banko et al., 2007; Banko et al., 2008) • Korean Treebank and Syntactic Heuristics  Wikipedia-based Approach (Wu and Weld, 2010) • Korean Wikipedia articles and Infoboxes Approach P R F Heuristic-based 92.31 17.27 29.09 Wikipedia-based 66.67 66.91 66.79 Projection-based 67.69 87.41 76.30 22
  • 23. Contents • Introduction • Methods  Cross-lingual Annotation Projection for Relation Extraction  Graph-based Projection Approach • Evaluation • Conclusions 23
  • 24. Conclusion • Summary  A graph-based projection approach for relation extraction • Label propagation algorithm • On a graph that represents the instance and context features of both the source and target languages  Experimental results show that our approach helps to improve the performances of relation extraction compared to other approaches • Future work  To relieve the high complexity problem of the approach  To deal with more expanded graph structure to improve the extraction performances 24
  • 25. Q&A