A Cross-Lingual ILP Solution to Zero Anaphora Resolution Ryu Iida & Massimo Poesio (ACL-HLT 2011)

Slides:



Advertisements
Similar presentations
Specialized models and ranking for coreference resolution Pascal Denis ALPAGE Project Team INRIA Rocquencourt F Le Chesnay, France Jason Baldridge.
Advertisements

A Machine Learning Approach to Coreference Resolution of Noun Phrases By W.M.Soon, H.T.Ng, D.C.Y.Lim Presented by Iman Sen.
Coreference Based Event-Argument Relation Extraction on Biomedical Text Katsumasa Yoshikawa 1), Sebastian Riedel 2), Tsutomu Hirao 3), Masayuki Asahara.
Jointly Identifying Temporal Relations with Markov Logic Katsumasa Yoshikawa †, Sebastian Riedel ‡, Masayuki Asahara †, Yuji Matsumoto † † Nara Institute.
Large-Scale Entity-Based Online Social Network Profile Linkage.
A Corpus for Cross- Document Co-Reference D. Day 1, J. Hitzeman 1, M. Wick 2, K. Crouch 1 and M. Poesio 3 1 The MITRE Corporation 2 University of Massachusetts,
1 A Description Logic with Concrete Domains CS848 presentation Presenter: Yongjuan Zou.
Beyond NomBank: A Study of Implicit Arguments for Nominal Predicates Matthew Gerber and Joyce Y. Chai Department of Computer Science Michigan State University.
Overview of the KBP 2013 Slot Filler Validation Track Hoa Trang Dang National Institute of Standards and Technology.
Chapter 18: Discourse Tianjun Fu Ling538 Presentation Nov 30th, 2006.
Reference Resolution #1 CSCI-GA.2590 Ralph Grishman NYU.
Predicting Text Quality for Scientific Articles Annie Louis University of Pennsylvania Advisor: Ani Nenkova.
Predicting Text Quality for Scientific Articles AAAI/SIGART-11 Doctoral Consortium Annie Louis : Louis A. and Nenkova A Automatically.
Detecting Anaphoricity and Antecedenthood for Coreference Resolution Olga Uryupina Institute of Linguistics, RAS.
Predicting the Semantic Orientation of Adjective Vasileios Hatzivassiloglou and Kathleen R. McKeown Presented By Yash Satsangi.
CS 4705 Lecture 21 Algorithms for Reference Resolution.
Page 1 Generalized Inference with Multiple Semantic Role Labeling Systems Peter Koomen, Vasin Punyakanok, Dan Roth, (Scott) Wen-tau Yih Department of Computer.
Empirical Methods in Information Extraction - Claire Cardie 자연어처리연구실 한 경 수
Supervised models for coreference resolution Altaf Rahman and Vincent Ng Human Language Technology Research Institute University of Texas at Dallas 1.
Improving Machine Learning Approaches to Coreference Resolution Vincent Ng and Claire Cardie Cornell Univ. ACL 2002 slides prepared by Ralph Grishman.
Extracting Opinions, Opinion Holders, and Topics Expressed in Online News Media Text Soo-Min Kim and Eduard Hovy USC Information Sciences Institute 4676.
A Global Relaxation Labeling Approach to Coreference Resolution Coling 2010 Emili Sapena, Llu´ıs Padr´o and Jordi Turmo TALP Research Center Universitat.
A Light-weight Approach to Coreference Resolution for Named Entities in Text Marin Dimitrov Ontotext Lab, Sirma AI Kalina Bontcheva, Hamish Cunningham,
Andreea Bodnari, 1 Peter Szolovits, 1 Ozlem Uzuner 2 1 MIT, CSAIL, Cambridge, MA, USA 2 Department of Information Studies, University at Albany SUNY, Albany,
Empirical Methods in Information Extraction Claire Cardie Appeared in AI Magazine, 18:4, Summarized by Seong-Bae Park.
C OLLECTIVE ANNOTATION OF WIKIPEDIA ENTITIES IN WEB TEXT - Presented by Avinash S Bharadwaj ( )
LREC 2010, Malta Maj Centre for Language Technology The DAD corpora and their uses Costanza Navarretta Funded by Danish Research.
2007. Software Engineering Laboratory, School of Computer Science S E Towards Answering Opinion Questions: Separating Facts from Opinions and Identifying.
Differential effects of constraints in the processing of Russian cataphora Kazanina and Phillips 2010.
Illinois-Coref: The UI System in the CoNLL-2012 Shared Task Kai-Wei Chang, Rajhans Samdani, Alla Rozovskaya, Mark Sammons, and Dan Roth Supported by ARL,
CROSSMARC Web Pages Collection: Crawling and Spidering Components Vangelis Karkaletsis Institute of Informatics & Telecommunications NCSR “Demokritos”
On the Issue of Combining Anaphoricity Determination and Antecedent Identification in Anaphora Resolution Ryu Iida, Kentaro Inui, Yuji Matsumoto Nara Institute.
Incorporating Extra-linguistic Information into Reference Resolution in Collaborative Task Dialogue Ryu Iida Shumpei Kobayashi Takenobu Tokunaga Tokyo.
Multi-modal Reference Resolution in Situated Dialogue by Integrating Linguistic and Extra-Linguistic Clues Ryu Iida Masaaki Yasuhara Takenobu Tokunaga.
1 Exploiting Syntactic Patterns as Clues in Zero- Anaphora Resolution Ryu Iida, Kentaro Inui and Yuji Matsumoto Nara Institute of Science and Technology.
A Bootstrapping Method for Building Subjectivity Lexicons for Languages with Scarce Resources Author: Carmen Banea, Rada Mihalcea, Janyce Wiebe Source:
1 Learning Sub-structures of Document Semantic Graphs for Document Summarization 1 Jure Leskovec, 1 Marko Grobelnik, 2 Natasa Milic-Frayling 1 Jozef Stefan.
Opinion Holders in Opinion Text from Online Newspapers Youngho Kim, Yuchul Jung and Sung-Hyon Myaeng Reporter: Chia-Ying Lee Advisor: Prof. Hsin-Hsi Chen.
Exploiting Context Analysis for Combining Multiple Entity Resolution Systems -Ramu Bandaru Zhaoqi Chen Dmitri V.kalashnikov Sharad Mehrotra.
COLING 2012 Extracting and Normalizing Entity-Actions from Users’ comments Swapna Gottipati, Jing Jiang School of Information Systems, Singapore Management.
Indirect Supervision Protocols for Learning in Natural Language Processing II. Learning by Inventing Binary Labels This work is supported by DARPA funding.
A Systematic Exploration of the Feature Space for Relation Extraction Jing Jiang & ChengXiang Zhai Department of Computer Science University of Illinois,
Summarization Focusing on Polarity or Opinion Fragments in Blogs Yohei Seki Toyohashi University of Technology Visiting Scholar at Columbia University.
An Entity-Mention Model for Coreference Resolution with Inductive Logic Programming Xiaofeng Yang 1 Jian Su 1 Jun Lang 2 Chew Lim Tan 3 Ting Liu 2 Sheng.
Minimally Supervised Event Causality Identification Quang Do, Yee Seng, and Dan Roth University of Illinois at Urbana-Champaign 1 EMNLP-2011.
Natural Language Programming David Vadas The University of Sydney Supervisor: James Curran.
Using Semantic Relations to Improve Passage Retrieval for Question Answering Tom Morton.
Multilingual Opinion Holder Identification Using Author and Authority Viewpoints Yohei Seki, Noriko Kando,Masaki Aono Toyohashi University of Technology.
Results of the 2000 Topic Detection and Tracking Evaluation in Mandarin and English Jonathan Fiscus and George Doddington.
Inference Protocols for Coreference Resolution Kai-Wei Chang, Rajhans Samdani, Alla Rozovskaya, Nick Rizzolo, Mark Sammons, and Dan Roth This research.
Number Sense Disambiguation Stuart Moore Supervised by: Anna Korhonen (Computer Lab)‏ Sabine Buchholz (Toshiba CRL)‏
Evaluation issues in anaphora resolution and beyond Ruslan Mitkov University of Wolverhampton Faro, 27 June 2002.
A Maximum Entropy Based Honorificity Identification for Bengali Pronominal Anaphora Resolution Apurbalal Senapati and Utpal Garain Presented by Samik Some.
Support Vector Machines and Kernel Methods for Co-Reference Resolution 2007 Summer Workshop on Human Language Technology Center for Language and Speech.
Improved Video Categorization from Text Metadata and User Comments ACM SIGIR 2011:Research and development in Information Retrieval - Katja Filippova -
1 Adaptive Subjective Triggers for Opinionated Document Retrieval (WSDM 09’) Kazuhiro Seki, Kuniaki Uehara Date: 11/02/09 Speaker: Hsu, Yu-Wen Advisor:
Finding document topics for improving topic segmentation Source: ACL2007 Authors: Olivier Ferret (18 route du Panorama, BP6) Reporter:Yong-Xiang Chen.
FILTERED RANKING FOR BOOTSTRAPPING IN EVENT EXTRACTION Shasha Liao Ralph York University.
A Statistical Model for Multilingual Entity Detection and Tracking R. Florian, H. Hassan, A. Ittycheriah, H. Jing, N. Kambhatla, X. Luo, N. Nicolov, S.
Identifying “Best Bet” Web Search Results by Mining Past User Behavior Author: Eugene Agichtein, Zijian Zheng (Microsoft Research) Source: KDD2006 Reporter:
Solving Hard Coreference Problems Haoruo Peng, Daniel Khashabi and Dan Roth Problem Description  Problems with Existing Coref Systems Rely heavily on.
Learning to Generate Complex Morphology for Machine Translation Einat Minkov †, Kristina Toutanova* and Hisami Suzuki* *Microsoft Research † Carnegie Mellon.
Part 2 Applications of ILP Formulations in Natural Language Processing
Simone Paolo Ponzetto University of Heidelberg Massimo Poesio
NYU Coreference CSCI-GA.2591 Ralph Grishman.
张昊.
Social Knowledge Mining
Location Recommendation — for Out-of-Town Users in Location-Based Social Network Yina Meng.
Clustering Algorithms for Noun Phrase Coreference Resolution
Extracting Why Text Segment from Web Based on Grammar-gram
Presentation transcript:

A Cross-Lingual ILP Solution to Zero Anaphora Resolution Ryu Iida & Massimo Poesio (ACL-HLT 2011)

Zero-anaphora resolution  Anaphoric function in which phonetic realization of anaphors is not required in “pro-drop” languages  Based on speaker and hearer’s shared understanding  φ : zero-anaphor (non-realized argument)  Essential: 64.3% of anaphors in Japanese newspaper articles are zeros (Iida et al. 2007) English: John went to visit some friends. On the way, he bought some wine. Italian: Giovanni andò a far visita a degli amici. Per via, φ comprò del vino. Japanese: John-wa yujin-o houmon-sita. Tochu-de φ wain-o ka-tta. 2

Research background  Zero-anaphora resolution has remained an active area for Japanese (Seki et al. 2002, Isozaki&Hirao 2003, Iida et al. 2007, Imamura et al. 2009, Sasano et al. 2009, Taira et al. 2010)  The availability of the annotated corpora such that provided by SemEVAL2010 task10 “Multi-lingual coreference (Recasens et al.2010) is leading to renewed interest (e.g. Italian)  Mediocre results obtained on zero anaphors by most systems in SemEVAL e.g. I-BART’s recall on zeros < 10% 3

Resolving zero-anaphors requires  The simultaneous decision of  Zero-anaphor detection: find phonetically unrealized arguments of predicates (e.g. verbs)  Antecedent identification: search for an antecedent of a zero-anaphor  Roughly correspond to anaphoricity determination and antecedent identification in coreference resolution  Denis&Baldridge(2007) proposed a solution to optimize the outputs from anaphoricity determination and antecedent identification by using Integer Linear Programming (ILP) 4

Main idea  Apply Denis&Baldridge (2007)’s ILP framework to zero-anaphora resolution  Extend the ILP framework into a two-way to make it more suitable for zero-anaphora resolution  Focus on Italian and Japanese zero-anaphora to investigate whether or not our approach is useful across languages  Study only subject zero-anaphors (only type in Italian) 5

Topic of contents  Research background  Denis&Baldridge (2007)’s ILP model  Proposal: extending the ILP model  Empirical evaluations  Summary & future directions 6

Denis&Baldrige (2007)’s ILP formulation of base model  object function  If, mentions i and j are coreferent and mention j is an anaphor : 1 if mentions i and j are coreferent; otherwise 0 7

Denis&Baldrige (2007)’s ILP formulation of joint model  object function  If, mentions i and j are coreferent and mention j is an anaphor; otherwise j is non-anaphoric : 1 if mention j is an anaphor; otherwise 0 8

3 constraints in ILP model characteristics of coreference relations transitivity of coreference chains 1. Resolve only anaphors: if mention pair ij is coreferent, mention j must be anaphoric 2. Resolve anaphors: if mention j is anaphoric, it must be coreferent with at least one antecedent 3. Do not resolve non-anaphors: if mention j is non-anaphoric, it should be have no antecedents 9

Proposal: extending the ILP framework  Denis&Baldridge’s original ILP-based model is not suitable for zero-anaphora resolution  Two modifications 1. Applying best-first solution 2. Incorporating a subject detection model 10

1. Best-first solution  Select at most one antecedent for an anaphor  “Do-not-resolve-anaphors” constraint is too weak Allow the redundant choice of more than one candidate antecedent Lead to decreasing precision on zero-anaphora resolution  “Do-not-resolve-anaphors” constraint is replaced with “Best First constraint (BF)” that blocks selection of more than one antecedent: 11

2. Integrating subject detection model  Zero-anaphor detection  Difficulty in zero-anaphora resolution comparing to pronominal reference resolution  Simply relying on the parser is not enough  most dependency parsers are not very accurate at identifying grammatical roles  detecting subject is crucial for zero-anaphor detection 12

2. Integrating subject detection model  Resolve only non-subjects: if a predicate j syntactically depends on a subject, the predicate j should have no antecedent of its zero anaphor : 1 if predicate j syntactically depends on a subject; otherwise 0 13

Experiment 1: zero-anaphors  Compare the baseline models with the extended ILP-based models  Use the Maximum Entropy model to create base classifiers in the ILP framework and baselines  Feature definitions basically follow the previous work (Iida et al. 2007) and (Poesio et al. 2010) 14

Two baseline models  PAIRWISE classification model (PAIRWISE)  Antecedent identification and anaphoricity determination are simultaneously executed by a single classifier (as in Soon et al. 2001)  Anaphoricity Determination-then-Search antecedent CASCADEd model (DS-CASCADE) 1. Filter out non-anaphoric candidate anaphors using an anaphoricity determination model 2. Select an antecedent from a set of candidate antecedents of anaphoric anaphors using an antecedent identification model 15

Data sets  Italian (Wikipedia articles)  LiveMemories text corpus 1.2 (Rodriguez et al. 2010) Data set on the SemEval2010: Coreference Resolution in Multiple Languages #zero-anaphors: train 1,160 / test 837  Japanese (newspaper articles)  NAIST text corpus (Iida et al. 2007) ver.1.4ß #zero-anaphors: train 29,544 / test 11,205 16

Creating subject detection models  Data sets  Italian: 80,878 tokens in TUT corpus (Bosco et al. 2010)  Japanese: 1753 articles (i.e. training dataset) in NAIST text corpus merged with Kyoto text corpus dependency arc is judged as positive if its relation is subject; as negative otherwise  Induce a maximum entropy classifier based on the labeled arcs  Features  Italian: lemmas, PoS tags and morphological information automatically computed by TextPro (Pianta et al. 2008)  Japanese: similar features as Italian except gender and number information 17

Results for zero anaphors ItalianJapanese modelRPFRPF PAIRWISE DS-CASCADE ILP ILP+BF ILP+SUBJ ILP+BF+SUBJ BF: use best first constraint, +SUBJ: use subject detection model 18

Experiment 2: all anaphors 19  Investigate performance of all anaphors (i.e. NP- coreference and zero-anaphors)  Use the same data set and same data separation  Italian: LiveMemories text corpus 1.2  Japanese: NAIST text corpus 1.4ß  Performance of each model are compared in terms of MUC score  Different types of referring expressions display very different anaphoric behavior  Induce 2 different models for NP-coreference and zero-anaphora respectively

ItalianJapanese modelRPFRPF PAIRWISE DS-CASCADE I-BART (Poesio et al. 2010) ILP ILP+BF ILP+SUBJ ILP+BF+SUBJ Results for all anaphors 20

Summary  Extended Denis&Baldridge (2007)’s ILP-based coreference resolution model by incorporating modified constraints & a subject detection model  Our results show the proposed model obtained improvement on both zero-anaphora resolution and overall coreference resolution 21

Future directions  Introduce more sophisticated antecedent identification model  Test our model for English constructions resembling zero-anaphora  Null instantiations in SEMEVAL 2010 ‘Linking Events and their Participants in Discourse’ task  Detect generic zero-anaphors  Have no antecedent in the preceding context  e.g. the Italian and Japanese translation of I walked into the hotel and (they) said … 22

23

Data sets on English coreference  Use ACE-2002 data set  Data set is classified into the two subset  Pronouns and NPs 24

Details of experiment: English 25 training data train: NPs train: zeros models: NP coreference models: zero anaphora test data test: NPs test: zeros outputs: all anaphors outputs: NPs outputs: zeros

Results: all anaphors (English) 26 English modelRPF PAIRWISE DS-CASCADE0.597 ILP ILP+BF ILP+SUBJ--- ILP+BF+SUBJ---