The Study on Automatic Chinese Collocation Extraction by Xu Ruifeng A Thesis Submitted in Partial Fulfillment of the Requirements _for the Degree of Doctor of Philosophy Department of Computing The Hong Kong Polytechnic University Jan 12, 2006 UMI Number: 3241091 INFORMATION TO USERS The quality of this reproduction is dependent upon the quality of the copy submitted. Broken or indistinct print, colored or poor quality illustrations and photographs, print bleed-through, substandard margins, and improper alignment can adversely affect reproduction. In the unlikely event that the author did not send a complete manuscript and there are missing pages, these will be noted. Also, if unauthorized copyright material had to be removed, a note will indicate the deletion.
® UMI UMI Microform 3241091 Copyright 2007 by ProQuest Information and Learning Company. All rights reserved. This microform edition is protected against unauthorized copying under Title 17, United States Code. ProQuest Information and Learning Company 300 North Zeeb Road P.
Box 1346 Ann Arbor, MI 48106-1346 CERTIFICATE OF ORIGINALITY I hereby declare that this thesis is my own work and that, to the best of my knowledge and belief, it reproduces no material previously published or written, nor material that has been accepted for the award of any other degree or diploma, except where due acknowledgement has been made in the text. Collocation is a lexical phenomenon in which two or more words are: habitually combined together as some conventional way of saying things. Collocation information is essential to many natural language processing tasks such as word sense disambiguation, machine translation, and information extraction. Most of current works on collocation extraction are statistical based with limited precision and recall because they can not well distinguish word co-occurrences, which are statistically significant, from true collocations, which are of habitual use and are thus either syntactically or semantically significant.
The objective of this study is to investigate methods to improve the performance of Chinese collocation extraction algorithms. Different types of collocations are identified. Collocation extraction algorithms are then desigried to target on different types of collocations using different features and criteria associated with these different types. In addition to improve statistical based collocation extraction algorithms, additional syntactic and semantic information are also incorporated into the algorithm to further improve the performance of collocation extraction.
On the study of the statistical based algorithms, a new algorithm based on bi-directional word bi-grams is designed to help identify collocations with low co-occurrence frequency and are of fixed use. A large scale Chinese collocation answer set is established so that collocation extraction algorithms can be evaluated and compared objectively by using the same training corpus and corresponding answer set. Collocations are then categorized into four types based on their compositionality, substitutability, modifiability and internal association. Based on the characteristics of each type of collocations, a multi-stage window- based collocation extraction is built where the n-gram collocations and different types of bi- gram collocations are separately extracted in different stages using different strategies and different discriminative features.
A Chinese shallow treebank, referred to as the PolyU Treebank, is annotated manually to provide syntactic and semantic knowledge to further help collocation extraction. This treebank is also used to train a chunker based on lexicalized Hidden Markov Model (HMM). The chunker provides ways to process running text for collocation extraction. By using the support collocation patterns and reject collocation patterns extracted from the annotated Chinese treebank and parsed running text, syntactic features are employed to further improve the performance of the window-based collocation extraction system.
Experimental results show that the use of syntactic patterns can significantly improve the - performance of collocation extraction, especially for filtering pseudo collocations. The extracted collocations were applied in the post-processing of a handwritten Chinese character recognition system. Experiments indicate that collocation information can be used in real application to improve the performance of these natural language related applications. Keyword: Collocation extraction, Treebank, Chunking and parsing, ii Publications Arising From the Thesis > [1] Guo-hong Fu, Ruifeng Xu, K.
Luke and Qin Lu, Chinese Text Chunking Using Lexicalized HMMs, In Proceedings of IEEE International Conference on Machine Learning and Cybernetics, pp.7-12, Guang Zhou, China, 2005 [2] Qin Lu, Jing Zhou and Ruifeng Xu, Machine Learning Approaches for Chinese Shallow Parsing, In Proceedings of IEEE International Conference on Machine Learning and Cybernetics, vol.2309-2314, Xi’an, China, 2003 {3] Qin Lu, S. Chan, Ruifeng Xu, et al. A Unicode Based Adaptive Segmentor, In Proceedings of 2nd Workshop on ACL SIGHAN, pp164-167, Sapporo, Japan, 2003 [4] Qin Lu, Yin Li and Ruifeng Xu, Improving Xtract for Chinese Collocation Extraction, In Proceedings of IEEE International Conference on Natural Language Processing and Knowledge Engineering, pp. 333-338, Beijing, China, 2003 [Š} Qin Lu, S.
Chan, Ruifeng Xu, et al. A Unicode Based Adaptive Segmentor, Journal of Chinese Language and Computing, vol. 3, pp221-234, 2004 [6] Ruifeng Xu, Qin Lu, Daniel S. Yeung and Xizhao Wang, Distant BI-Gram Model, Collocation and Their Application in Post-processing for Chinese Character Recognition, In Proceedings of IEEE International Conference on Machine Learning and Cybernetics, vol.2251-2255, Beijing, China, 2002 [7] Ruifeng Xu, Qin Lu and Yin Li, An Automatic Chinese Collocation Extraction Algorithm based on Lexical Statistics, In Proceedings of IEEE International Conference on Natural Language Processing and Knowledge Engineering, pp.321-326, Beijing, China, 2003 [8] Ruifeng Xu, Qin Lu and Wanyin Li, The Construction of a Chinese Shallow Treebank, In Proceedings of 3rd Workshop on ACL SIGHAN, pp.94-101,Barcelona, Spain, 2004 nỉ [9] Ruifeng Xu, Daniel Yeung and Daming Shi, A Hybrid Post-processing System for Offline Handwritten Chinese Character Recognition based on a Statistical Language Model.
International Journal of Pattern Recognition and Artificial Intelligence, vol. 415-428, 2005 ˆ [10] Ruifeng Xu, Qin Lu, Yin Li and Wanyin Li, The Design and Construction of the PolyU Shallow Treebank, International Journal of Computational Linguistics and Chinese Language Processing, vol.397-416, 2005 [11] Ruifeing Xu and Qin Lu, Multi-stage Chinese Collocation Extraction, In Proceedings ofIEEE International Conference on Machine Learning and Cybernetics, pp.3254-3259, Guang Zhou, China, 2005 | [12] Ruifeng Xu and Qin Lu, Improving Collocation Extraction by Using Syntactic Patterns, In Proceedings of IEEE Conference on Natural Language Processing and Knowledge Engineering 2005, pp.52-57, WuHan, China, 2005 [13] Ruifeng Xu and Qin Lu, A Multi-stage Collocation Extraction System, The Advances in Machine Learning and Cybernetics, Lecture Notes on Artificial Intelligences (LNAI 3930), (Yeung D.), Springer-Verlag, Berlin Heidelberg: pp.740-749, 2006 [14] Ruifeng Xu, Qin Lu and Sujian Li, The Design and Construction of a Chinese Collocation Bank, Accpeted to published in Proceedins of Ffth International Conference on Language Resources and Evaluation, Genoa, Italy, 2006 [15] SuJian Li, Wen-jie Li, Qin Lu and Ruifeng Xu, Verifying Person Descriptions with Term-Entity Association, In Proceedings ofIEEE International Conference on Machine Learning and Cybernetics, pp. 50-55, Guang Zhou, China, 2005 [16] Sujian Li, Yun Li, Luning Ji and Ruifeng Xu, Use of Dictionary Matching and String Frequency Statistics in Content Analysis of Automatic Indexing, In Proceedings of 8th Joint Symposium on Computational Linguistics, NanJing, China, 2005 [17] Wanyin Li, Qin Lu and Ruifeng Xu, Using Synonym Relations in Chinese Collocation Extraction, In Proceedings of 3rd Workshop on ACL SIGHAN, pp.86-93, Barcelona, Spain, 2004 [18] Wanyin Li, Qin Lu and Ruifeng Xu, Similarity based Chinese Synonyms Collocation Extraction, International Journal of Computational Linguistics and Chinese Language Processing, vol. This thesis could not have been done without the help and cooperation of many peoples, and it is now my great pleasure to take this opportunity to thank them.
First and foremost, I would like to express my deepest thanks to my supervisor, Prof. Qin Lu, for being a consistent source of support and encouragement. I could not imagine having a better advisor and mentor for me. Without her knowledge and perceptiveness, I would never have finished my Ph.
I gratefully acknowledge her who gives me enormous freedom to pursue my own interests while at the same time providing just the right amount of guidance to ensure the right research approach. It would be my great pleasure to thank Dr. James Liu, Prof. William Wong and Prof.
Maosong Sun, for their great efforts, valuable comments and excellent advices to improve the quality and readability of the earlier version of this thesis. I would like to thank Dr. WenJie Li, who constantly encouraged me, contributed her valuable insight in academic research during the past three years. Another great excellent person whom I would like to express my deep gratitude is Dr.
Sujian Li, my close friend, for the continuous support and kindly help. I would also like to thank all my friends in our research group, Mrs. Yin Li, Ms. WanYan Li, Mr.
Tin-Shing Chiu, Ms. LuNing Ji, Mr. Ming-Li Wu, Dr. BaoLi Li, Mr.
Wei Li and Mr. Qing Chen, who have always been the great support to me and have made this group a wonderful place to learn and have fun. I will treasure their friendship for the rest of my life. At last, I would like express my deepest appreciation to my father Yu-Shu Xu and my mother Shao-Lan Li, my wife Shu-Qi Jiang, my aunt Shu-Jun Xu and my uncle Bao-Ku Su, vii my elder brother Rui-Song Xu and his wife Di An, for their endless love and unwavering support.
This thesis is dedicated to my mother Shao-Lan Li who gave my live, to my wife Shu-Qi rd Jiang who gives me true love, and to my baby daughter Carol Xu, the hope. vi Table of Contents Certificate of Originality À0.nn i Publications Asing from the Th€SI§. c0 ng TH cọ 4 64 0 10 11 FC VN ố. vii Table of Contents.
ix List Of Figures .sccccccsccscssssnssesscescnccssssneeseessssessssesecescesessoasseeasscaeasoesseeseeeseesssneseusaeseaseosees xili ID 82g 1. -- -‹- ‹ sóc sọ Họ Họ họ 9 0 th 1 2 Basic Concepts and Thesis SCOD€.2 Motivation and Problem Sta†€Tn€T(. Research Objectives and Thesis Scope. Gc HQ KH 6 0.1 Review of Automatic Collocation Extraction Techn1que€s.1 Window-based Statistical Collocation Extraction Approach.2 Syntax-based Collocation Extraction Approach.3 Collocation Extraction using Semantic ÌnÍOrmatiOH.2 Review of Automatic Shallow P4TSIDE.1 Statistic-based Shallow PaTSiDE.2 Rule-based Shallow P4TSINE.- - --c cm HH mờ 35 4 Collocation Extraction Based on Lexical Statistics.1 Preparation of Training Corpus and Answer S€(.- cọ HH nh 40 4.
Applying Xtract to Chinese Collocation Extraction: CX?r4cf. Improving CXtract: CXÍTđCÍÏ.4 A New Collocation Extraction System: CX?rC£2.1 The Framework DeSIØH.- - ---- cà cà ng HH4 08000842 157 52 4.2 Construct a Word Co-occurrence Database for CXtract2 ¬ 57 4.3 Evaluation Of CX?rđC(2.5 Evaluations of Statistical Collocation Extraction Algorithms.-- so + HH4 ng tình ng nh ni, 64 5 Multi-Stage Collocation ExtractiOn.1 Categorization of Chinese CollocatIOnS.--Ặ sen ng g4 th re.2 Characteristic Analysis of Typical CollocatiOnS. The Design of A Multi-Stage Collocation Extraction System.1 Additional Feature Selections .2 Applying the Heuristic Rules to Eliminate Pseudo Collocations.3 The New Multi-stage Extraction Algorithm.4 Parameter Optimization based on Perceptron Training Rule.4 Experimental Results and EvaluatiOTiS.--- cà Sen 421118821 188823 xe 86 5.1 Experimental Data PreparatOTn.2 Experiments on Type 1 and Type 2 Collocation Extraction in Stage 3 .3 Experiments on Weight Parameter Ôptimization.4 Experiments on Multi-stage Collocation Extraction of Stage 1-3 and 5.5 Experiments on Pseudo collocation Filtering by Using Heuristic Rules.6 Experiments on Evaluating the Complete Collocation Extraction System 93 5.5 Chapter Sumưnar1z2tiOT.- - -- S2 - SH ki n0 9v 94 6 The Design and Development of Chinese Shallow Treebank and Automatic Chunkers 95 6.1 The Design and Development of PolyU Treebank .1 Basic Concepts and Background of Shallow Treebank .--- - ------ HH HH th K1 0011 1088124 28791 99 6.3 Annotation Guideline Desigm.-- - -- - --- SH HH HH nen 102 6.4 Implementation of the PolyU Treebank.- co SSS+ĂẰ he enreg 108 6.5 Quality Assurance and Annotation PTOBT€SS.6 Contributions of PolyU Treebank.2 The Design and Development of Automatic ChuniK€TS.1 Chunking Scope and R€pr€s€ntatiOn.- nen ng errườ 118 6.2 Chunking with POS FeatUr€s.-- -- HH HH HH ng 120 6.3 Chunking with Lexicalized FeafUT€S. -- - -- - Ác HH ng gey 125 6.4 Experiments and EvaluUatiOTS.