Recognizing Textual Entailment Using
Total Page:16
File Type:pdf, Size:1020Kb
RECOGNIZING TEXTUAL ENTAILMENT USING DESCRIPTION LOGIC AND SEMANTIC RELATEDNESS Reda Siblini A thesis in The Department of Computer Science and Software Engineering Presented in Partial Fulfillment of the Requirements For the Degree of Doctor of Philosophy Concordia University Montreal,´ Quebec,´ Canada April 2014 c Reda Siblini, 2014 Concordia University School of Graduate Studies This is to certify that the thesis prepared By: Mr. Reda Siblini Entitled: Recognizing Textual Entailment using Description Logic and Semantic Relatedness and submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy (Computer Science) complies with the regulations of this University and meets the accepted standards with respect to originality and quality. Signed by the final examining committee: Dr. A. Hammad, Chair Dr. H. Saggion, External Examiner Dr. L. Kosseim, Thesis Supervisor Dr. V. Haarslev, Examiner Dr. S. Bergler, Examiner Dr. F. Khendek, Examiner Approved Chair of Department or Graduate Program Director 20 Dr. Christopher Trueman, Dean Faculty of Engineering and Computer Science Abstract Recognizing Textual Entailment using Description Logic and Semantic Relatedness Reda Siblini, Ph.D. Concordia University, 2014 Textual entailment (TE) is a relation that holds between two pieces of text where one read- ing the first piece can conclude that the second is most likely true. Accurate approaches for textual entailment can be beneficial to various natural language processing (NLP) ap- plications such as: question answering, information extraction, summarization, and even machine translation. For this reason, research on textual entailment has attracted a signif- icant amount of attention in recent years. A robust logical-based meaning representation of text is very hard to build, therefore the majority of textual entailment approaches rely on syntactic methods or shallow semantic alternatives. In addition, approaches that do use a logical-based meaning representation, require a large knowledge base of axioms and inference rules that are rarely available. The goal of this thesis is to design an efficient description logic based approach for recognizing textual entailment that uses semantic re- latedness information as an alternative to large knowledge base of axioms and inference rules. In this thesis, we propose a description logic and semantic relatedness approach to textual entailment, where the type of semantic relatedness axioms employed in aligning the descrip- tion logic representations are used as indicators of textual entailment. In our approach, the text and the hypothesis are first represented in description logic. The representations are enriched with additional semantic knowledge acquired by using the web as a corpus. The iii hypothesis is then merged into the text representation by learning semantic relatedness axioms on demand and a reasoner is then used to reason over the aligned representation. Finally, the types of axioms employed by the reasoner are used to learn if the text entails the hypothesis or not. To validate our approach we have implemented an RTE system named AORTE, and evaluated its performance on recognizing textual entailment using the fourth recognizing textual entailment challenge. Our approach achieved an accuracy of 68.8% on the two way task and 61.6% on the three way task which ranked the approach as 2nd when compared to the other participating runs in the same challenge. These results show that our description logical based approach can effectively be used to recognize textual entailment. iv Acknowledgments This thesis reflects my long and rewarding research journey, during which I came across many wonderful people to whom I express my thanks. First, I would like to express my deepest graduate to my supervisor, Dr. Leila Kos- seim, whose spirit of quest in regard to research inspired me to freely explore new ideas. This thesis would not have been possible without her persistent support and guidance. In addition, I wish to extend my appreciation for the members of my doctoral commit- tee, Professors Volker Haarslev, Sabine Bergler, Ferhat Khendek, and Amin Hammad for their valuable time and feedback. I am particularly thankful to Professor Bergler, whose enthusiasm for linguistics is contagious, for her thorough comments on my the- sis. I am also grateful to Professor Haarslev, who introduced me and gave me access to his work in Description Logic. I want to specially thank the external examiner, Professor Horacio Saggion for his participation in my defense committee and for his valuable feedback on my thesis. Finally, I would like to thank my colleagues at the CLaC Lab, my family, and loved ones for their continuous moral support and encouragement throughout my studies. v Contents List of Figures x List of Tables xii 1 Introduction 1 1.1 Problem Statement . 2 1.2 Motivation . 5 1.3 Overview of the Thesis . 7 1.4 Intended Contributions . 9 1.5 Thesis Organization . 11 2 State of the Art 13 2.1 Approaches to Recognizing Textual Entailment . 14 2.1.1 Lexical Based Methods . 15 2.1.2 Syntax Based Methods . 18 2.1.3 Semantic Based Methods . 20 2.1.4 Logical Form Based Methods . 21 2.1.5 Hybrid Methods . 23 2.1.6 Analysis . 24 2.2 Recognizing Textual Entailment Benchmarks . 26 2.2.1 The First Recognizing Textual Entailment Challenge (RTE1) . 26 vi 2.2.2 The Second Recognizing Textual Entailment Challenge (RTE2) 31 2.2.3 The Third Recognizing Textual Entailment Challenge (RTE3) 32 2.2.4 The Fourth Recognizing Textual Entailment Challenge (RTE4) 33 2.2.5 The Fifth Recognizing Textual Entailment Challenge (RTE5) 35 2.2.6 The Student Response Analysis and The Eighth Recognizing Textual Entailment Challenge (SRA-RTE8) . 36 2.2.7 Analysis of Benchmarking . 37 2.3 Resources used in Recognizing Textual Entailment . 38 2.3.1 Acronym Lists . 39 2.3.2 Nominalization Databases . 40 2.3.3 Gazetteers . 42 2.3.4 N-gram Models . 42 2.3.5 Semantic Networks . 43 2.3.6 Frames . 46 2.3.7 Thesauri and Encyclopaedias . 49 2.3.8 Ontologies . 51 2.3.9 Inference Rules . 53 2.4 Conclusion . 55 3 Baseline Knowledge Querying Approach 56 3.1 Overview of the Baseline Knowledge Querying Approach . 57 3.2 Representation . 61 3.2.1 Syntactic Analysis . 65 3.2.2 Semantic Analysis . 70 3.2.3 Ontological Analysis . 74 3.3 Comparison . 84 3.3.1 Query Creation . 85 vii 3.3.2 Querying with a Reasoner . 88 3.4 Decision . 90 3.5 Evaluation and Analysis . 91 3.5.1 Evaluation of the Knowledge Querying Approach . 91 3.5.2 Evaluation of the Recognizing Textual Entailment Approach . 93 3.5.3 Analysis . 94 3.6 Conclusion . 95 4 Knowledge Alignment Approach 96 4.1 Overview of the Knowledge Alignment Approach . 97 4.2 Representation . 101 4.2.1 Named Entity Recognition Using the Web . 101 4.2.2 Evaluation of the Named Entity Recognizer . 107 4.2.3 Named Entity Recognizer and RTE . 110 4.3 Comparison . 111 4.4 Decision . 116 4.5 Evaluation and Analysis . 120 4.6 Conclusion . 123 5 Recognizing Textual Entailment and Lexical Semantic Relatedness 124 5.1 Lexical Semantic Relatedness . 125 5.1.1 Our Approach to Semantic Relatedness . 127 5.2 Intrinsic Evaluation of Semantic Relatedness . 133 5.2.1 Evaluation using Human Ratings . 133 5.2.2 Evaluation using Synonym Tests . 136 5.2.3 Evaluation on a Word-Phrase Semantic Relatedness . 138 5.3 Knowledge Alignment Approach to RTE Revisited . 140 viii 5.4 Evaluation and Analysis . 143 5.5 Conclusion . 144 6 Conclusion and Future Work 146 6.1 Main Findings and Contributions of the Thesis . 147 6.2 Directions for Future Research . 150 Bibliography 152 A Sample OWL Generated Representation 178 ix List of Figures 1 Baseline Knowledge Querying Approach . 59 2 Prototype of the Baseline Knowledge Querying Approach . 60 3 Syntax and Semantics of SHOINQ Description Logic from [Horrocks and Patel-Schneider, 2004] . 63 4 A Graphical Illustration of Part of the Representation Created for the Text \Jurassic Park is a novel written by Michael Crichton and Pub- lished in 1990." with WordNet axioms. 66 5 Syntax and Semantics of OWL DL Entities [Horrocks, Patel-Schneider, and Van Harmelen, 2003] . 77 6 Syntax and Semantics of OWL DL Axioms and Facts [Horrocks, Patel- Schneider, and Van Harmelen, 2003] . 78 7 A Graphical Illustration of Part of the Representation Created for the Text \Jurassic Park is a novel written by Michael Crichton and pub- lished in 1990.".............................. 80 8 Knowledge Alignment Approach . 98 9 Prototype of the Knowledge Alignment Approach . 99 10 A Graphical Illustration of representation-T Created for the Hypothe- sis \Jurassic Park is a novel written by Michael Crichton" . 114 11 A Graphical Illustration of representation-H Created for the Hypothe- sis \Michael Crichton crafted the book Jurassic Park" . 115 x 12 A Graphical Illustration of an Alignment representation-A . 116 13 Part of the Resulting Decision Tree . 119 14 Example of the Semantic Network Around the Word car. 128 15 Lowest Cost Path Between the Words Monk and Oracle. 132 16 Knowledge Alignment Approach Augmented with New Semantic Re- latedness Calculator . 142 xi List of Tables 1 Dependency Parse of \Jurassic Park is a novel written by Michael Crichton and Published in 1990.".................... 67 2 Result of the Syntactic Analysis for \Jurassic Park is a novel written by Michael Crichton and Published in 1990. 69 3 Result of the Semantic Analysis for the \Jurassic Park is a novel writ- ten by Michael Crichton and Published in 1990. 74 4 Result of the Semantic Analysis for \Michael Crichton crafted Jurassic Park in 1990." .............................. 85 5 Examples of semantic mapping for the predicate structure has-crafter (Craft, Author) .............................. 87 6 Examples of Predicate Structure Patterns and Corresponding nRQL Queries . 89 7 Sample of the Evaluation Results .