臺灣博碩士論文加值系統

文字蘊涵是自然語言處理最近興起的研究課題。文字蘊涵識別(Recognizing Textual Entailment, RTE)可以應用到其他許多自然語言處理的研究中。在本文中將介紹我們在觀察NTCIR-10-RITE-2資料集後發現過去系統的缺陷，進而提出如何改進中文文字蘊涵系統的方法。過去的系統處理文字蘊涵都使用機器學習分類文題的方法，所有輸入句子都用同樣的分類器處理，對於某些特別的問題往往會產生誤判。我們認為應該針對於特定類型的問題做處理，增加系統可以處理的問題類型。實驗結果顯示配合之前提出的機器學習方法，增加四種特殊類型分類對特殊類型句子進行個別處理，可以有效改進系統，實驗結果系統在識別簡體中文蘊涵兩類的正確率從原本67.86%提昇到72.62%。

Recognizing Textual Entailment (RTE) is a new research issue in natural language processing (NLP) research area. RTE can be a useful component in many NLP applications. In this paper, we introduce our finding on the entailment analysis of the NTCIR-10 RITE-2 dataset, and use the observation to improve our system. In the previous works, all the input pairs are treated equally in a standard classification architecture. We find that is not suitable for some special cases. We believe that by isolating the special cases and building separated classifiers, a RTE system can perform better. After implementing modules for four special cases into our system, the result is significantly improved from 67.86% to 72.62% on the binary class classification task.

目錄
摘要 I
ABSTRACT II
銘謝 III
目錄 V
表目錄 VII
圖目錄 VIII
第一章 緒論 1
第二章　相關研究 3
2.1中文文字蘊涵推論研究 5
2.2其他語言研究 6
2.3 過去研究總結 7
第三章　語料集與資料處理 11
3.1中文語料集 11
3.2資料處理 13
3.2.1格式正規化 13
3.2.2背景知識的替換 14
第四章　系統架構與實驗結果 16
4.1使用工具 17
4.2系統子單元 18
4.2.1基於特徵分類法 18
4.2.2特殊類型處理 24
4.3實驗結果 35
4.3.1實驗一 baseline系統 35
4.3.2實驗二加入單語言機器翻譯特徵 35
4.3.3實驗三加入是否詞規則特徵 35
4.3.4實驗四測試各特徵對系統效能影響 36
4.3.5 實驗五特殊類型處理-肯定/否定句 37
4.3.6實驗六 特殊類型處理-時間資訊不一致 38
4.3.7實驗七 特殊類型處理-數字資訊不一致 39
4.3.8 實驗八 特殊類型處理-主/受詞資訊不同 39
4.3.9 實驗總結 40
五、結論與未來工作 42
六、參考文獻 45

表目錄
表一. 各種蘊涵關係例句 4
表二 SHIH[9]等人提出使用三個分類器的真值表 8
表三 SHIH[12]等人基於語言學上的方法分類表 9
表四 NTCIR-10 RITE-2[13]日文任務單元處理測試列表 10
表五 中文語料集分兩類統計 12
表六 中文語料集分兩類統計 12
表七 各種時間表示方法例句[5] 13
表八 數字正規化例句 14
表九 系統使用的特徵列表 18
表十 基於字元數特徵值計算例子 19
表十一 基於BIEU特徵值計算例子 21
表十二 GIZA++匹配得分列表 23
表十三 容易誤判問題類型例子 25
表十四 特殊類型例子 30
表十五 訓練集與測試集特殊類型數量統計 34
表十六BASELINE系統簡體與繁體中文實驗結果 35
表十七BASELINE加入單一語言機器翻譯實驗結果 35
表十八 實驗二加入是否詞規則特徵實驗結果 36
表十九 測試各特徵對系統效能影響實驗結果 37
表二十 特殊類型處理-肯定/否定句與實驗三比較結果 38
表二十一 特殊類型處理-時間資訊不一致與實驗三比較結果 38
表二十二 特殊類型處理-時間資訊不一致與實驗三比較結果 39
表二十三 特殊類型處理-主/受詞資訊不同與實驗三比較結果 39
表二十四 實驗總結表 40
表二十五 NTCIR-10 RITE-2[24]各隊簡體中文兩類最佳成果 41

圖目錄
圖一 WANG[8]等人提出反向的矛盾字詞對齊示意圖 8
圖二 分兩類語料集 12
圖三 系統流程圖 16
圖四 句子字詞匹配示意圖 23

[1]Ido Dagan, Oren Glickman and Bernardo Magnini.The PASCAL Recognising Textual Entailment Challenge.In Quinonero-Candela, J.; Dagan, I.; Magnini, B.; d''Alche-Buc, F. (Eds.) Machine Learning Challenges. Lecture Notes in Computer Science , Vol. 3944, pp. 177-190, Springer, 2006.
[2]Ido Dagan and Oren Glickman, Probabilistic textual entailment: Generic applied modeling of language variability, In Proceedings of the Workshop on Learning Methods for Text Understanding and Mining, Grenoble, France, 2004.
[3]Yongping Ou, Changqing Yao, “Recognize Textual Entailment by the Lexical and Semantic Matching”, Computer Application and System Modeling, 2010 International Conference on V2-500 -504
[4]Dong-Bin Hua, Jun Dinga,” Study on Similar Engineering Decision Problem Identification Based on Combination of Improved Edit-Distance and Skeletal Dependency Tree with POS”, Systems Engineering Procedia Volume 1, 2011, Pages 406–413.
[5]Shih-Hung Wu, Wan-Chi Huang, Liang-Pu Chen and Tsun Ku. Binary-class and Multi-class Chinese Textural Entailment System Description in NTCIR-9 RITE, in Proceedings of the NTCIR-9 workshop, Tokyo, Japan, 6-10 Dec., 2011.
[6]Yaoyun Zhang, Jun Xu,Chenlong Liu, Xiaolong Wang, RuifengXu,Qingcai Chen, Xuan Wang, YongshuaiHou and Buzhou Tang. “ICRC_HITSZ at RITE: Leveraging Multiple Classifiers Voting for Textual Entailment Recognition”, in NTCIR-9 RITE, in Proceedings of the NTCIR-9 workshop, Tokyo, Japan, 6-10 Dec., 2011
[7]Shan-Shun Yang, Shih-Hung Wu, Liang-Pu Chen, Wen-Tai Hsieh, and Seng-cho T. Chou, Improving Binary-class Chinese Textural Entailment by Monolingual Machine Translation Technology, in Proceedings of the IEEE IRI 2012, Las Vegas, USA,8 Aug, 2012.
[8]Xiao-Lin Wang, Hai Zhao, and Bao-Liang Lu, “BCMI-NLP Labeled-Alignment-Based Entailment System for NTCIR-10 RITE-2 Task”, in Proceedings of the NTCIR-10 conference, Tokyo, Japan, 18-21 June., 2013.
[9] Cheng-Wei Shih, Chad Liu, Cheng-Wei Lee,”IASL RITE System at NTCIR-10”,in Proceedings of the NTCIR-10 conference, Tokyo, Japan, 18-21 June., 2013.
[10] Chuan-Jie Lin and Bo-Yu Hsiao, “The Description of the NTOU RITE System in NTCIR-9”,in Proceedings of the NTCIR-9 workshop, Tokyo, Japan, 6-10 Dec., 2011
[11]Chun Tu , Min-Yuh Day,”Chinese Textual Entailment with Wordnet Semantic and Dependency SyntacticAnalysis”,in Proceedings of the IEEE IRI 2013,San Francisco, USAAug 14, 2013.
[12] Cheng-Wei Shih, Cheng-Wei Lee, Richard Tzonghan Tsai, Wen-Lian Hsu, “Validating Contradiction in Texts Using Online Co-MentionPattern Checking”,ACM Transactions on Asian Language Information Processing, Vol. 11, No. 4, Article 17, Publication date: December 2012.
[13]Yotaro Watanabe, Yusuke Miyao, Junta Mizuno, Tomohide Shibata, Hiroshi Kanayama, Cheng -Wel Lee, Chuan-Jie Lin , Shuming Shi, Teruko Mitamura, Noriko Kando, Hideki Shima, Kohichi Takeda,” Overview of the Recognizing Inference in Text (RITE-2)at the NTCIR-10 Workshop”, in Proceedings of the NTCIR-10 conference, Tokyo, Japan, 18-21 June., 2013.
[14]Qun Liu , Sujian Li, Word Similarity Computing Based on How-net, Computational Linguistics and Chinese Language Processing，Vol.7, No.2, August 2002, pp.59-76
[15]Min-Hsiang Li, Shih-Hung Wu, Yi-Ching Zeng, Ping-che Yang, and Tsun Ku, Chinese Characters Conversion System based on Lookup Table and Language Model, Computational Linguistics and Chinese Language Processing, Vol. 15, No. 1, March 2010, pp. 19-36.
[16]Kishore Papineni, Salim Roukos, Todd Ward and Wei-Jing Zhu, “BLEU: a method for automatic evaluation of machine translation”, In Proceedings of the 40th Annual Meeting on ACL, pages 311–318, Philadelphia, PA, 2002.
[17]Liang Zhou, Chin-Yew Lin and Eduard Hovy, “Re-evaluating machine translation results with paraphrase support”, In Proceedings of the Conference on EMNLP, pages 77–84, Sydney, Australia, 2006.
[18]Franz Josef Och and Hermann Ney, “A systematic comparison of various statistical alignment models.”Computational Linguistics, vol. 29, no. 1, pp. 19–51, 2003.
[19] M. Quang, Nhat Pham; Le Minh Nguyen; Shimazu, A. “Using Machine
Translation for Recognizing Textual Entailment in Vietnamese Language”, 2012 IEEE RIVF International Conference on Computing and Communication Technologies, Research, Innovation, and Vision for the Future .
[20] ICTCLAS, http://ictclas.org/
[21] CKIP, http://ckipsvr.iis.sinica.edu.tw
[22] LIBSVM, http://www.csie.ntu.edu.tw/~cjlin/libsvm/
[23] Stanford parser , http://nlp.stanford.edu/software/lex-parser.shtml
[24] GIZA++,http://www.statmt.org/moses/giza/GIZA++.html
[25] NTCIR-10 RITE-2,
http://www.cl.ecei.tohoku.ac.jp/rite2/doku.php