| { |
| "paper_id": "O05-1026", |
| "header": { |
| "generated_with": "S2ORC 1.0.0", |
| "date_generated": "2023-01-19T07:58:53.040463Z" |
| }, |
| "title": "", |
| "authors": [], |
| "year": "", |
| "venue": null, |
| "identifiers": {}, |
| "abstract": "", |
| "pdf_parse": { |
| "paper_id": "O05-1026", |
| "_pdf_hash": "", |
| "abstract": [], |
| "body_text": [ |
| { |
| "text": "\u5982\u300c E i E i+1 \u6b63\u78ba\u5c0d\u61c9 C j C j+1 C j+2 \u300d\u5373\u8868\u793a\u4e0d\u8ad6\u662f E i E i+1 \u6216 C j C j+1 C j+2 \u7686\u7121\u6cd5\u518d\u5207\u5272\u4ee5\u5f97 \u5230\u66f4\u5c0f\u7684\u6b63\u78ba\u5c0d\u61c9\u3002\u90e8\u4efd\u6b63\u78ba\u5c0d\u61c9\u4ee5\u4e0a\u8ff0\u6b63\u78ba\u5c0d\u61c9\u70ba\u4f8b\uff0c \u4efb\u610f {E i , E i+1 } \u7684\u5b50\u96c6\u5408\u5c0d\u61c9\u4efb\u610f {C j , C j+1 , C", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "", |
| "sec_num": null |
| } |
| ], |
| "back_matter": [], |
| "bib_entries": { |
| "BIBREF0": { |
| "ref_id": "b0", |
| "title": "Aligning sentences in parallel corpora", |
| "authors": [ |
| { |
| "first": "F", |
| "middle": [], |
| "last": "Peter", |
| "suffix": "" |
| }, |
| { |
| "first": "Jennifer", |
| "middle": [ |
| "C" |
| ], |
| "last": "Brown", |
| "suffix": "" |
| }, |
| { |
| "first": "Robert", |
| "middle": [ |
| "L" |
| ], |
| "last": "Lai", |
| "suffix": "" |
| }, |
| { |
| "first": "", |
| "middle": [], |
| "last": "Mercer", |
| "suffix": "" |
| } |
| ], |
| "year": 1991, |
| "venue": "Meeting of the Association for Computational Linguistics", |
| "volume": "", |
| "issue": "", |
| "pages": "169--176", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Peter F. Brown, Jennifer C. Lai, and Robert L. Mercer. Aligning sentences in parallel corpora. In Meeting of the Association for Computational Linguistics, pages 169-176, 1991.", |
| "links": null |
| }, |
| "BIBREF1": { |
| "ref_id": "b1", |
| "title": "Deriving translation data from bilingual texts", |
| "authors": [ |
| { |
| "first": "R", |
| "middle": [], |
| "last": "Catizone", |
| "suffix": "" |
| }, |
| { |
| "first": "G", |
| "middle": [], |
| "last": "Russell", |
| "suffix": "" |
| }, |
| { |
| "first": "S", |
| "middle": [], |
| "last": "Warwick", |
| "suffix": "" |
| } |
| ], |
| "year": 1989, |
| "venue": "Proceedings of the First Lexical Acquisition Workshop", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "R. Catizone, G. Russell, and S. Warwick. Deriving translation data from bilingual texts. Proceedings of the First Lexical Acquisition Workshop, 1989.", |
| "links": null |
| }, |
| "BIBREF2": { |
| "ref_id": "b2", |
| "title": "Extraction of translation unit from chinese-english parallel corpora", |
| "authors": [ |
| { |
| "first": "B", |
| "middle": [], |
| "last": "Chang", |
| "suffix": "" |
| }, |
| { |
| "first": "P", |
| "middle": [], |
| "last": "Danielsson", |
| "suffix": "" |
| }, |
| { |
| "first": "W", |
| "middle": [], |
| "last": "Teubert", |
| "suffix": "" |
| } |
| ], |
| "year": 2002, |
| "venue": "COLING-02: The First SIGHAN Workshop on Chinese", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "B. Chang, P. Danielsson, and W. Teubert. Extraction of translation unit from chinese-english parallel corpora. COLING-02: The First SIGHAN Workshop on Chinese, 2002.", |
| "links": null |
| }, |
| "BIBREF3": { |
| "ref_id": "b3", |
| "title": "A new approach for aligning parallel texts. COLING-94: 15th International Conference on Computational Linguistics", |
| "authors": [ |
| { |
| "first": "P", |
| "middle": [], |
| "last": "Fung", |
| "suffix": "" |
| }, |
| { |
| "first": "K", |
| "middle": [], |
| "last": "Church", |
| "suffix": "" |
| } |
| ], |
| "year": 1994, |
| "venue": "", |
| "volume": "", |
| "issue": "", |
| "pages": "1096--1102", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "P. Fung and K. Church. K-vec: A new approach for aligning parallel texts. COLING-94: 15th International Conference on Computational Linguistics, pages 1096-1102, Aug 1994.", |
| "links": null |
| }, |
| "BIBREF4": { |
| "ref_id": "b4", |
| "title": "Aligning noisy parallel corpora across language groups: Word pair feature matching by dynamic time warping", |
| "authors": [ |
| { |
| "first": "P", |
| "middle": [], |
| "last": "Fung", |
| "suffix": "" |
| }, |
| { |
| "first": "K", |
| "middle": [], |
| "last": "Mckeown", |
| "suffix": "" |
| } |
| ], |
| "year": 1994, |
| "venue": "", |
| "volume": "", |
| "issue": "", |
| "pages": "81--88", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "P. Fung and K. McKeown. Aligning noisy parallel corpora across language groups: Word pair feature matching by dynamic time warping. AMTA-94, Association for Machine Translation in the Americas, pages 81-88, 1994.", |
| "links": null |
| }, |
| "BIBREF5": { |
| "ref_id": "b5", |
| "title": "A program for aligning sentences in bilingual corpora", |
| "authors": [ |
| { |
| "first": "W", |
| "middle": [], |
| "last": "Gale", |
| "suffix": "" |
| }, |
| { |
| "first": "K", |
| "middle": [], |
| "last": "Church", |
| "suffix": "" |
| } |
| ], |
| "year": 1991, |
| "venue": "Proceedings of the Annual Conference of the Association for Computational Linguistics", |
| "volume": "", |
| "issue": "", |
| "pages": "177--184", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "W. Gale and K. Church. A program for aligning sentences in bilingual corpora. Proceedings of the Annual Conference of the Association for Computational Linguistics, pages 177-184, 1991.", |
| "links": null |
| }, |
| "BIBREF6": { |
| "ref_id": "b6", |
| "title": "Text-translation alignment", |
| "authors": [ |
| { |
| "first": "M", |
| "middle": [], |
| "last": "Kay", |
| "suffix": "" |
| }, |
| { |
| "first": "M", |
| "middle": [], |
| "last": "R\u00f6scheisen", |
| "suffix": "" |
| } |
| ], |
| "year": 1993, |
| "venue": "Computational linguistics", |
| "volume": "", |
| "issue": "", |
| "pages": "121--142", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "M. Kay and M. R\u00f6scheisen. Text-translation alignment. Computational linguistics, (1):121-142, 1993.", |
| "links": null |
| }, |
| "BIBREF7": { |
| "ref_id": "b7", |
| "title": "Clause alignment for hong kong legal texts: A lexical-based approach", |
| "authors": [ |
| { |
| "first": "C", |
| "middle": [], |
| "last": "Kit", |
| "suffix": "" |
| }, |
| { |
| "first": "J", |
| "middle": [ |
| "J" |
| ], |
| "last": "Webster", |
| "suffix": "" |
| }, |
| { |
| "first": "K-K", |
| "middle": [], |
| "last": "Sin", |
| "suffix": "" |
| }, |
| { |
| "first": "H", |
| "middle": [], |
| "last": "Pan", |
| "suffix": "" |
| }, |
| { |
| "first": "H", |
| "middle": [], |
| "last": "Li", |
| "suffix": "" |
| } |
| ], |
| "year": 2004, |
| "venue": "International Journal of Corpus Linguistics", |
| "volume": "", |
| "issue": "1", |
| "pages": "29--51", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "C. Kit, J. J. Webster, K-K. Sin, H. Pan, and H. Li. Clause alignment for hong kong legal texts: A lexical-based approach. International Journal of Corpus Linguistics, (1):29-51, 2004.", |
| "links": null |
| }, |
| "BIBREF8": { |
| "ref_id": "b8", |
| "title": "Models of co-occurrence", |
| "authors": [ |
| { |
| "first": "I", |
| "middle": [ |
| "D" |
| ], |
| "last": "Melamed", |
| "suffix": "" |
| } |
| ], |
| "year": 1998, |
| "venue": "IRCS Technical Report", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "I. D. Melamed. Models of co-occurrence. IRCS Technical Report, 1998.", |
| "links": null |
| }, |
| "BIBREF9": { |
| "ref_id": "b9", |
| "title": "Models of translational equivalence", |
| "authors": [ |
| { |
| "first": "I", |
| "middle": [ |
| "D" |
| ], |
| "last": "Melamed", |
| "suffix": "" |
| } |
| ], |
| "year": 2000, |
| "venue": "Computational Linguistics", |
| "volume": "", |
| "issue": "", |
| "pages": "221--249", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "I. D. Melamed. Models of translational equivalence. Computational Linguistics, pages 221-249, 2000.", |
| "links": null |
| }, |
| "BIBREF10": { |
| "ref_id": "b10", |
| "title": "Association-based bilingual word alignment", |
| "authors": [ |
| { |
| "first": "Robert", |
| "middle": [ |
| "C" |
| ], |
| "last": "Moore", |
| "suffix": "" |
| } |
| ], |
| "year": 2005, |
| "venue": "Proceedings, Workshop on Building and Using Parallel Texts: Data-Driven Machine Translation and Beyond", |
| "volume": "", |
| "issue": "", |
| "pages": "1--8", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Robert C. Moore. Association-based bilingual word alignment. Proceedings, Workshop on Building and Using Parallel Texts: Data-Driven Machine Translation and Beyond, Ann Arbor, Michigan, pages 1-8, 2005.", |
| "links": null |
| }, |
| "BIBREF11": { |
| "ref_id": "b11", |
| "title": "Aligning a parallel english-chinese corpus statistically with lexical criteria", |
| "authors": [ |
| { |
| "first": "D", |
| "middle": [], |
| "last": "Wu", |
| "suffix": "" |
| } |
| ], |
| "year": 1994, |
| "venue": "Proceedings of the 32nd Annual meeting of the Association for Computational Linguistics", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "D. Wu. Aligning a parallel english-chinese corpus statistically with lexical criteria. Proceedings of the 32nd Annual meeting of the Association for Computational Linguistics, 1994.", |
| "links": null |
| }, |
| "BIBREF12": { |
| "ref_id": "b12", |
| "title": "Subsentential translation memory for computer assisted writing and translation", |
| "authors": [ |
| { |
| "first": "Jian-Cheng", |
| "middle": [], |
| "last": "Wu", |
| "suffix": "" |
| }, |
| { |
| "first": "Thomas", |
| "middle": [ |
| "C" |
| ], |
| "last": "Chuang", |
| "suffix": "" |
| }, |
| { |
| "first": "Wen-Chi", |
| "middle": [], |
| "last": "Shei", |
| "suffix": "" |
| }, |
| { |
| "first": "Jason", |
| "middle": [ |
| "S" |
| ], |
| "last": "Chang", |
| "suffix": "" |
| } |
| ], |
| "year": 2004, |
| "venue": "The Companion Volume to the Proceedings of 42st Annual Meeting of the Association for Computational Linguistics", |
| "volume": "", |
| "issue": "", |
| "pages": "106--109", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Jian-Cheng Wu, Thomas C. Chuang, Wen-Chi Shei, and Jason S. Chang. Subsentential translation memory for computer assisted writing and translation. In The Companion Volume to the Proceed- ings of 42st Annual Meeting of the Association for Computational Linguistics, pages 106-109, Barcelona, Spain, July 2004. Association for Computational Linguistics.", |
| "links": null |
| }, |
| "BIBREF13": { |
| "ref_id": "b13", |
| "title": "Automatic alignment of english-chinese bilingual texts of cns news. Computational Linguistic Archive", |
| "authors": [ |
| { |
| "first": "D", |
| "middle": [], |
| "last": "Xu", |
| "suffix": "" |
| }, |
| { |
| "first": "C", |
| "middle": [ |
| "L" |
| ], |
| "last": "Tan", |
| "suffix": "" |
| } |
| ], |
| "year": 1996, |
| "venue": "", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "D. Xu and C. L. Tan. Automatic alignment of english-chinese bilingual texts of cns news. Compu- tational Linguistic Archive, 1996.", |
| "links": null |
| } |
| }, |
| "ref_entries": { |
| "TABREF0": { |
| "num": null, |
| "text": "\u8a5e\u5c0d\u61c9\u6f14\u7b97\u6cd5 Fung \u8207 Church (1994) \u7684 K-vec \u6f14\u7b97\u6cd5\u5c07\u96d9\u8a9e\u8a9e\u6599\u5eab\u5404\u5207\u5206\u70ba\u76f8\u7b49\u7684 K \u5340\u584a\uff0c \u6bcf\u4e00\u8a5e\u7686\u8a18 \u9304\u8a72\u8a5e\u5728 K \u500b\u5340\u584a\u4e2d\u51fa\u73fe\u8207\u5426\uff0c \u7d44\u6210\u4e00 K \u7dad vector (v 1 , v 2 , . . . , v K )\uff0cv i \u2208 {0, 1}\u3002 \u5c0d\u96d9\u8a9e\u7684 \u5169\u5169\u8a5e\u5f59\uff0c\u7686\u900f\u904e\u5f7c\u6b64 vector \u8a08\u7b97\u5404\u81ea\u983b\u7387\u53ca\u5171\u540c\u51fa\u73fe\u65bc\u76f8\u540c\u5340\u584a\u7684\u983b\u7387\uff0c \u4e26\u4ee5 MI (Mutual Information) \u4f86\u8a08\u7b97\u5169\u8a5e\u5f59\u7684\u76f8\u4f9d\u7a0b\u5ea6\u3002 \u7531\u65bc MI \u5c0d\u65bc\u983b\u7387\u751a\u5c11\u7684\u8a5e\u6703\u8a08\u7b97\u51fa\u6975\u5927\u503c\uff0c\u56b4\u91cd\u5f71\u97ff \u53ef\u4fe1\u5ea6\uff0c \u56e0\u6b64\u85c9\u7531 t-score \u503c\u4fee\u6b63\uff0c\u900f\u904e\u7d66\u5b9a\u7684\u5e38\u6578\u503c\uff0c\u5ffd\u7565 t-score \u503c\u5c0f\u65bc\u8a72\u5e38\u6578\u503c\u7684\u7d50\u679c\uff0c \u5c07\u5927\u5927\u63d0\u6607 MI \u7684\u53ef\u4fe1\u5ea6\u3002 \u7531\u65bc K-vec \u6f14\u7b97\u6cd5\u9700\u8981\u5207\u5272\u96d9\u8a9e\u8a9e\u6599\u6210 K \u500b\u5340\u584a\uff0c\u932f\u8aa4\u7684\u5207\u5272\u5c07\u4f7f\u7d50\u679c\u4e0d\u5982\u9810\u671f\u3002 \u56e0 \u6b64 Fung \u8207 McKeown (1994) \u518d\u63d0\u51fa DK-vec \u4ee5\u89e3\u6c7a\u6b64\u554f\u984c\u3002 \u5728 DK-vec \u4e2d\u6bcf\u4e00\u8a5e\u5f59\u7686\u8a18\u9304\u5169 vector\uff0c position vector \u8a18\u9304\u8a72\u8a5e\u5f59\u51fa\u73fe\u65bc\u96d9\u8a9e\u8a9e\u6599\u7684\u6240\u6709\u4f4d\u7f6e\uff0c recency vector \u5247\u8a18\u9304\u5169\u5169\u4f4d\u7f6e \u7684\u8ddd\u96e2\u3002 \u4ee5 position vector \u7684\u8cc7\u6599\u70ba\u6a6b\u5ea7\u6a19\u503c\uff0crecency vector \u7684\u8cc7\u6599\u70ba\u7e31\u5ea7\u6a19\u503c\uff0c \u4e26\u9023\u63a5\u76f8\u9130 \u4e4b\u9ede\uff0c\u5247\u53ef\u5f97\u4e00\u5206\u4f48\u65bc 2-D \u5ea7\u6a19\u7cfb\u7684\u51fd\u5f0f\u5206\u4f48\u53d6\u6a23\u5716\u3002 \u5229\u7528 pattern matching \u7684 Dynamic Time Gale \u8207Church (1991) \u53ca Brown \u7b49 (1991) \u65b9\u6cd5\u4e00\u6a23\uff0c Kay and R\u00f6scheisen (1993) \u5247\u5982 Kit et al. (2004) \u4ee5\u96d9\u8a9e\u6cd5\u5f8b\u689d\u6587\u7684 glossary \u548c\u96d9\u8a9e\u8fad\u5178\uff0c \u518d\u52a0\u4e0a\u9069\u7576\u7684\u6a19\u9ede\u7b26\u865f\u8f49\u63db\u3001 \u6578\u5b57\u8f49\u63db (\u5982\u963f\u62c9\u4f2f\u6578\u5b57\u8207\u7f85\u99ac\u6578\u5b57)\uff0c\u518d\u8a2d\u8a08\u4e00\u4f30\u8a08\u51fd\u6578\u4f86\u7d50\u5408 \u5168\u90e8\u8cc7\u8a0a\u800c\u5f97\u76f8\u4f3c\u7a0b\u5ea6\uff0c \u4ee5\u5176\u8a55\u4f30\u5b57\u53e5\u5c0d\u61c9\uff0c\u53ef\u9054 94.6% \u7684\u6b63\u78ba\u7387\u3002 Kit et al. (2004) \u4ee5\u8a5e\u5f59\u8a0a Association-based binlingual word alignment \u4e2d\uff0c \u8a5e\u5f59\u7684\u51fa\u73fe\u983b\u7387\u626e\u6f14\u8457\u95dc\u9375\u7684\u89d2\u8272\u3002 \u4e0d \u8ad6\u4f7f\u7528 MI\u3001t-score \u6216\u8005 LLR \u4f86\u8a55\u4f30\u5169\u8a5e\u5f59\u7684\u76f8\u4f9d\u7a0b\u5ea6\uff0c \u7686\u5229\u7528\u983b\u7387\u7684\u8cc7\u8a0a\u4f86\u8a08\u7b97\u3002 \u800c\u53e6\u4e00\u95dc \u8868\u793a\u4e00\u82f1\u6587 (\u4e2d\u6587) \u5340\u584a\uff0c \u7a31\u6b64\u6642\u7684\u5207\u5206\u72c0\u614b\u70ba \u2126\u3002 \u4ee4 B e i = e i,1 e i,2 . . .\uff0cB c j = c j,1 c j,2 . . .\uff0c \u5176\u4e2d e i,k (c j,l ) \u8868\u793a\u4e00\u82f1\u6587 (\u4e2d\u6587) \u8a5e\u5f59\u3002 \u4ee4 asso(e, c) \u8868\u793a\u8a5e\u5f59 e \u548c\u8a5e\u5f59 c \u7684\u76f8\u4f9d\u6b0a\u503c\u5927\u5c0f (\u6b64\u76f8\u4f9d\u6b0a\u503c\u53ef\u8996\u9700\u8981\u9078\u7528\u5982 MI\u3001t-score\u3001LLR \u7b49\u3002 \u5728\u672c \u5be6\u9a57\u4e2d\u6211\u5011\u4ee5 MI \u70ba\u4e3b\uff0c\u642d\u914d t-score \u4ee5\u904e\u6ffe\u8a5e\u983b\u4f4e\u7684\u5c0d\u61c9)\uff0c \u5247 ASSO(\u2126) = \u2211 \u5373\u70ba\u5728 \u2126 \u5207\u5206\u72c0\u614b\u4e0b\u7684\u7e3d\u9ad4\u76f8\u4f9d\u503c\u3002 \u4ee4 new(\u2126, i, start e , end e , start c , end c ) \u8868\u793a\u4e00\u7a2e\u65b0\u7684\u5207\u5206 \u72c0\u614b\uff0c \u5176\u610f\u7fa9\u70ba\u5728 \u2126 \u5207\u5206\u72c0\u614b\u4e2d\uff0c\u7b2c i \u5340\u584a\u88ab\u5207\u5206\u4e86\uff0c \u5207\u5206\u65b9\u5f0f\u70ba e i,starte e i,starte+1 . . . e i,ende \u548c c i,startc c i,startc+1 . . . c i,endc \u70ba \u4e00 \u7d44 \u5c0d \u61c9 \u5340 \u584a \uff0c \u800c e i,1 e i,2 . . . e i,starte\u22121 e i,ende+1 . . . e i,|Ei| \u548c c i,1 c i,2 . . . c i,startc\u22121 c i,endc+1 . . . c i,|Ci| \u70ba\u53e6\u4e00\u7d44\u5c0d\u61c9\u5340\u584a\u3002 \u56e0\u6b64\uff0c\u5c0d \u2126 \u72c0\u614b\u800c\u8a00\uff0c\u8a08\u7b97 ASSO(new(\u2126, i, start e , end e , start c , end c )) \u2200i = 1, 2, . . . , |\u2126| \u82e5 value > ASSO(\u2126)\uff0c\u5373\u8868\u793a\u8a72\u5207\u5206\u65b9\u5f0f\u80fd\u5920\u63d0\u9ad8\u7e3d\u9ad4\u76f8\u4f9d\u503c\uff0c \u4f9d\u6b64\u6642\u4e4b start e \u3001 end e \u3001 start c \u3001 end c \u9032\u884c\u5207\u5206\uff0c \u53ef\u5f97\u4e00\u65b0\u7684\u5207\u5206\u72c0\u614b \u2126 \u3002\u82e5 value \u2264 ASSO(\u2126)\uff0c \u8868\u793a\u6240\u6709\u7684\u5207\u5206\u65b9 \u5728\u76ee\u524d\u5207\u5206\u72c0\u614b \u2126 \u4e2d\uff0c\u5c0d\u6bcf\u4e00\u5340\u584a\u9032\u884c\u5207\u5206\u5617\u8a66\uff0c\u4e26\u8a18\u9304\u65b0\u7684\u5207\u5206\u65b9\u5f0f\u65bc \u2126 \u3002 3. \u5982\u679c |\u2126| = |\u2126 | \u5247\u7d50\u675f\uff0c\u5426\u5247\u56de\u5230 2.\u3002 \u52a0\u901f\u8207\u5be6\u4f5c \u5728\u4e0a\u8ff0\u6f14\u7b97\u6cd5\u4e2d\uff0c\u7531\u65bc\u8981\u5c0d\u6240\u6709\u53ef\u80fd\u7684\u5207\u5206\u65b9\u5f0f\u8a08\u7b97 ASSO \u503c\uff0c \u4ea6\u5373\u5c0d\u65bc\u6240\u6709\u53ef\u80fd\u7684\u5207\u5206 \u65b9\u5f0f\u90fd\u8981\u57f7\u884c\u4e00\u6b21\u985e\u4f3c K-vec \u7684\u6f14\u7b97\u904e\u7a0b\uff0c \u5247\u6b64\u6f14\u7b97\u6cd5\u7684\u8a08\u7b97\u8907\u96dc\u5ea6\u5c07\u6703\u5341\u5206\u5730\u9ad8\uff0c \u5728\u5be6\u4f5c\u4e0a \u96d6\u7136\u4e26\u4e0d\u56f0\u96e3\uff0c\u4f46\u8a08\u7b97\u6642\u9593\u5c07\u6703\u5341\u5206\u5730\u4e45\u3002 \u800c\u82e5\u662f\u52a0\u4e0a\u5c0d start e|c \u53ca end e|c \u7684\u9650\u5236\uff0c\u5c07\u6703\u6709\u6548 \u6070\u7d04\u7b49\u65bc\u7e3d\u8a5e\u6578 (\u4e2d\u6587 3291\u3001\u82f1\u6587 3908)\uff0c \u56e0\u6b64\u6211\u5011\u4ee5\u6bb5\u843d\u4f5c\u70ba K-vec \u7684\u5340\u584a\u3002 \u76f8\u4f9d\u6b0a\u503c\u4f7f\u7528 MI \u53ca t-score \u5224\u5225\uff0c MI \u53ca t-score \u7684\u53c3\u6578\u6bd4\u7167\u539f\u4f5c\u8005\u7684\u5efa\u8b70\uff1a\u4ee5 t-score \u503c\u70ba\u7be9\u9078\u5668\uff0c \u53ea\u8003\u616e t-score \u2265 1.65 \u7684\u8a5e\u5c0d\u61c9\u3002MI \u5247\u505a\u70ba\u4e3b\u8981\u76f8\u4f9d\u6b0a\u503c\u7684\u4f9d\u64da\uff0c \u8f38\u51fa\u6642 \u4ee5 MI \u7684\u503c\u7531\u5927\u5230\u5c0f\u6392\u5e8f\uff0c\u4e26\u6368\u68c4 MI < 1.0 \u7684\u7d50\u679c\u3002 \u5728\u9019\u500b\u689d\u4ef6\u4e0b\u7684\u8f38\u51fa\u5982\u4e0b\u8868\u6240\u793a\uff1a \u6f14\u7b97\u6cd5 t-score \u7be9\u9078\u503c MI \u6700\u5c0f\u503c \u8a5e\u5c0d\u61c9\u6578 \u6b63\u78ba\u6578 precision \u96d6\u7136\u5728 MI \u2265 1.0 \u7684\u8f38\u51fa\u689d\u4ef6\u4e0b\uff0c\u6211\u5011\u7684\u6f14\u7b97\u6cd5 precision \u8f03\u4f4e\uff0c \u4f46\u7531 Figure1 \u548c Figure 2 \u53ef \u770b\u51fa\u5728\u540c\u6a23\u500b\u6578\u7684\u8f38\u51fa (\u8f38\u51fa\u7686\u4ee5 MI \u7684\u6b0a\u503c\u5927\u5c0f\u70ba\u5e8f)\u4e0b\uff0c\u6211\u5011\u7684\u6f14\u7b97\u6cd5\u6709\u8f03\u597d\u7684\u8868\u73fe\u3002 \u5728\u8f38 \u51fa\u524d 10 \u689d\u8a5e\u5c0d\u61c9\u6642\uff0c\u6211\u5011\u7684\u6f14\u7b97\u6cd5\u548c K-vec \u5dee\u7570\u4e0d\u5927\uff0c \u4f46\u5f9e\u7b2c 10 \u689d\u8a5e\u5c0d\u61c9\u4e4b\u5f8c\u7684\u8f38\u51fa\u7d50\u679c\uff0c \u660e\u986f\u6211\u5011\u7684\u6f14\u7b97\u6cd5\u6709\u66f4\u9ad8\u7684\u6b63\u78ba\u7387\uff0c \u5230\u524d 30 \u9805\u8f38\u51fa\u4ecd\u7dad\u6301 0.6 \u4ee5\u4e0a\u7684 precision\uff0c \u800c\u5728\u524d 30 \u9805 \u8f38\u51fa\u6642 K-vec \u7684 precision \u50c5\u7d04 0.45\u3002", |
| "content": "<table><tr><td>2 \u76f8\u95dc\u7814\u7a76 \u7684\u8a5e) \u53ca\u5728\u6587\u7ae0\u4e2d\u51fa\u73fe\u7684\u5206\u4f48\uff0c \u5efa\uf9f7\u53ef\u80fd\u7684\u8a5e\u5c0d\u61c9\u8868\u53ca\uf906\u5c0d\u61c9\u8868\u4e26\uf967\u65b7\u7684\u4fee\u6b63\uff0c \u4ee5 relaxation \u65b9 \u5f0f\u90fd\u7121\u6cd5\u518d\u63d0\u9ad8\u7e3d\u9ad4\u76f8\u4f9d\u6b0a\u503c\uff0c\u56e0\u6b64\u8a72\u5340\u584a\u6c92\u6709\u518d\u88ab\u5207\u5206\u7684\u5fc5\u8981\u3002 \u91cd\u8986\u6b64\u4e00\u6b65\u9a5f\uff0c\u5247\u5207\u5206\u4e4b\u5340 \u53ef\u77e5\uff0c\u5728\u73fe\u884c\u689d\u4ef6\u4e0b\uff0cS e maxe \u5c0d\u61c9 S c maxc \u662f\u6700\u53ef\u4fe1\u8cf4\u7684\u3002 \u56e0\u6b64\uff0c\u5c07 S e maxe \u8207 S c maxc \u53d6\u51fa\u4f7f\u6210\u65b0 \u7531\u4e0b\u9762\u5217\u8209\u7684\u8a5e\u5c0d\u61c9\u7d50\u679c\u53ef\u4ee5\u770b\u51fa\uff0c\u90e8\u4efd\u6b63\u78ba\u7684\u8a5e\u5c0d\u61c9\u4f54\u4e86\u76f8\u7576\u7684\u6bd4\u4f8b\uff0c \u5982\u679c\u52a0\u4e0a\u9019\u4e9b\u8907 \u53e5\u6578 \u53e5\u5e73\u5747\u8a5e\u6578 \u6a19\u6e96\u5dee</td></tr><tr><td>1 \u5c0e\u8a00 \u8a9e\u8a00\u7ffb\u8b6f\u5728\u8cc7\u8a0a\u7684\u50b3\u905e\u4e0a\u626e\u6f14\u5341\u5206\u91cd\u8981\u7684\u89d2\u8272\uff0c\u5728\u904e\u53bb\uff0c\u8a9e\u8a00\u7ffb\u8b6f\u7684\u5de5\u4f5c\u7686\u4ee5\u4eba\u5de5\u7ffb\u8b6f\u70ba \u4e3b\u3002 \u7531\u65bc\u96fb\u8166\u79d1\u5b78\u7684\u9032\u6b65\uff0c\u904b\u7b97\u80fd\u529b\u5927\u5e45\u63d0\u9ad8\uff0c\u5404\u985e\u76f8\u95dc\u7684\u7406\u8ad6\u3001\u6f14\u7b97\u6cd5\u4e5f\u76f8\u7e7c\u88ab\u63d0\u51fa\uff0c \u5982\u4f55 \u5229\u7528\u96fb\u8166\u4f86\u9032\u884c\u81ea\u52d5\u7ffb\u8b6f\u5de5\u4f5c\u6210\u4e86\u91cd\u8981\u7684\u7814\u7a76\u8ab2\u984c\u3002 \u5728\u8a31\u591a\u81ea\u52d5\u7ffb\u8b6f\u7684\u7814\u7a76\u4e2d\uff0c\u8a5e\u5c0d\u61c9 (word alignment) \u662f\u4e0d\u53ef\u6216\u7f3a\u7684\u91cd\u8981\u6b65\u9a5f\uff0c \u5176\u6b63\u78ba\u7387\u5f80\u5f80\u5c0d\u7ffb\u8b6f\u7684\u7d50\u679c\u6709\u95dc\u9375\u6027\u7684\u5f71\u97ff\u3002 \u50b3\u7d71\u8a5e\u5c0d\u61c9 \u4e43\u662f\u7531\u4eba\u5de5\u6240\u5efa\u7acb\uff0c\u5982\u96d9\u8a9e\u8a5e\u5178\u5373\u662f\u4eba\u5de5\u5efa\u7acb\u7684\u8a5e\u5c0d\u61c9\u8cc7\u6599\u5eab\u3002 \u4f46\u4eba\u5de5\u5efa\u7acb\u4e0d\u4f46\u8cbb\u6642\u8cbb\u529b\uff0c\u96e3 \u4ee5\u8ddf\u4e0a\u65b0\u8a5e\u589e\u52a0\u7684\u901f\u5ea6\uff0c \u4e14\u8a5e\u5178\u6709\u5176\u6975\u9650\uff0c\u518d\u5b8c\u5584\u7684\u8a5e\u5178\u7686\u4e0d\u53ef\u80fd\u5305\u542b\u6240\u6709\u96d9\u8a9e\u8a5e\u5f59\u5c0d\u61c9\u3002 \u52a0 \u4ee5\u73fe\u4eca\u7db2\u8def\u4e0a\u5df2\u6709\u5927\u91cf\u7684\u96d9\u8a9e\u6a5f\u8b80\u8cc7\u6599\uff0c\u5728\u7814\u7a76\u8cc7\u6599\u5145\u6c9b\u7684\u60c5\u5f62\u4e0b\uff0c \u7531\u96fb\u8166\u81ea\u52d5\u5efa\u7acb\u8a5e\u5c0d\u61c9\u4ea6 \u6cd5\u9054\u5230\u6536\u6582\u3002 \u8207 \u7684\u65b9\u6cd5\u53ea\u6709\u5728\u4e00\u5c0d\u4e00\u7684\u60c5\u5f62\u4f54\u7d55\u5927\u591a\uf969\u6642\u624d\u6703\u6709\u597d\u7684\u6548\u679c\u3002 \u6b64\u5916\u6b64\u7a2e\u65b9\u6cd5\u904e\ufa01\u91cd\u8996\u8a5e\u983b\uff0c\u6587\u7ae0 \u7684\u9577\ufa01\u592a\u77ed\u6703\u9020\u6210\u6b63\u78ba\uf961\u7684\u5927\u5e45\u4e0b\ufa09\u3002 \u9019\u500b\u6f14\u7b97\u6cd5\u53e6\u4e00\u500b\u5be6\u505a\u4e0a\u7684\u554f\u984c\u662f\u8655\uf9e4\u5341\u5206\u8017\u6642\uff0c\u7121\u6cd5 \u5feb\u901f\u8655\uf9e4\u5927\uf97e\u8a9e\uf9be\u3002 \u5b50\u53e5\u5c0d\u61c9 (clause alignment) \u606f\u5f97\u5230\u975e\u5e38\u9ad8\u7684\u5c0f\u53e5\u5c0d\u61c9\u6b63\u78ba\u7387\u7684\u4e3b\u8981\u539f\u56e0\u662f\u6240\u7528\u7684\u8a9e\u6599\u70ba\u6cd5\u5f8b\u96d9\u8a9e\u6587\u4ef6\u4e14\u4f7f\u7528\u6cd5\u5f8b\u8853\u8a9e\u7684\u8fad \u5178\uff0c \u4e14\u6b64\u985e\u6587\u4ef6\u4e2d\u4ee3\u8868\u6cd5\u5f8b\u689d\u6587\u7684\u6578\u5b57\u4e00\u518d\u51fa\u73fe\u3002 \u5728\u6211\u5011\u4e4b\u524d\u7684\u5be6\u9a57 (\u6797\u8207\u9ad8 (2004) ) \u986f\u793a\u5728 \u4e00\u822c\u7684\u4e2d\u82f1\u96d9\u8a9e\u6587\u7ae0\u4f7f\u7528\u96d9\u8a9e\u8fad\u5178\u3001\u6578\u5b57\u3001 \u53ca\u6a19\u9ede\u8a0a\u606f\u5728\u5927\u53e5\u7684\u5c0d\u61c9\u6b63\u78ba\u7387\u5c1a\u4e14\u4e0d\u5230 90%\uff0c \u5c0f \u53e5\u7684\u6b63\u78ba\u7387\u5fc5\u5b9a\u7121\u6cd5\u9054\u5230 Kit et al. (2004) \u7684\u6c34\u6e96\u3002 \u584a\u6578\u5c07\u6703\u4e0d\u65b7\u589e\u52a0\uff0c \u76f4\u5230\u6240\u6709\u7684\u5340\u584a\u90fd\u7121\u6cd5\u518d\u88ab\u5207\u5206\u3002 \u6f14\u7b97\u6cd5\u5982\u4e0b\uff1a 1. \u4ee5\u96d9\u8a9e\u8a9e\u6599\u7684\u6bb5\u843d\u5c0d\u61c9\u4f5c\u70ba\u521d\u59cb\u5207\u5206\u72c0\u614b\u3002 2. 4. \u5229\u7528\u76ee\u524d\u5207\u5206\u72c0\u614b\u6c42\u51fa\u8a5e\u5f59\u9593\u7684\u76f8\u4f9d\u6b0a\u503c\u4e26\u8f38\u51fa\u7d50\u679c\u3002 3.2 \u6bb5\u843d\u5c0d\u9f4a\u5e73\u884c\u8a9e\u6599\u7684\u8a5e\u5c0d\u61c9\u66a8\u5c0f\u53e5\u5c0d\u61c9\u6f14\u7b97\u6cd5 \u7684\u5340\u584a\u3002 \u5c0d\u65bc\u6bcf\u4e00\u5340\u584a\uff0c\u91cd\u8986\u9019\u500b\u6b65\u9a5f\uff0c\u76f4\u5230\u5340\u584a\u7684\u7e3d\u6578\u4e0d\u518d\u8b8a\u52d5\u3002 asso(e, c) \u4e43\u6839\u64da\u76ee\u524d\u70ba\u6b62\u7684\u8a5e\u5c0d\u61c9\u76f8\u4f9d\u6b0a\u503c\u4f86\u8a08\u7b97\uff1b \u800c\u63a5\u8457\u8a5e\u5c0d\u61c9\u4e43\u6839\u64da\u65b0\u7684\u5207\u5206\u72c0\u614b\u4f86 \u6c42\u5176\u76f8\u4f9d\u6b0a\u503c\u3002\u4ea4\u4e92\u8fed\u4ee3\u5f8c\u5c07\u6703\u6536\u6b5b\uff0c \u4ea6\u5373\u5169\u8005\u7686\u4e0d\u518d\u8b8a\u52d5\u3002\u7531\u65bc\u6211\u5011\u7684\u5207\u5206\u662f\u4ee5\u6a19\u9ede\u7b26\u865f\u5207 \u5206\u7684\u5b50\u5340\u584a\u70ba\u55ae\u4f4d\uff0c \u56e0\u6b64\u82e5\u76ee\u524d\u8655\u7406\u5340\u584a\u53ea\u5305\u542b\u4e00\u500b\u5b50\u5340\u584a\uff0c\u5247\u4e0d\u53ef\u518d\u88ab\u5340\u5206\uff0c \u56e0\u6b64\u8a72\u6f14\u7b97\u6cd5 \u4fdd\u8b49\u6703\u6536\u6b5b\u3002 \u800c\u521d\u59cb\u7684\u6bb5\u843d\u5c0d\u61c9\u5247\u7528\u65bc\u63d0\u4f9b\u6700\u521d\u7684\u76f8\u4f9d\u6b0a\u503c\u8a08\u7b97\uff0c \u6b64\u5916\u4e5f\u4fdd\u8b49\u521d\u59cb\u7684\u5340\u584a\u5c0d\u61c9 \u662f\u5b8c\u5168\u6b63\u78ba\u7684\u3002 4 \u5be6\u9a57\u6750\u6599 \u672c\u7814\u7a76\u4f7f\u7528\u7684\u4e2d\u82f1\u5c0d\u8b6f\u6587\u7ae0\u53d6\u81ea\u5149\u83ef\u96dc\u8a8c (http://www.sinorama.com.tw/ch/)\uff0c\u7d71\u8a08\u8cc7\u6599\u5982 \u5408\u8a5e\u7684\u64f7\u53d6\u6f14\u7b97\u6cd5\uff0c\u5247\u53ef\u671b\u5927\u5e45\u63d0\u5347 precision\u3002 \u6b64\u5916\uff0c\u7531\u65bc\u57fa\u65bc\u7d71\u8a08\u7684\u76f8\u4f9d\u6b0a\u503c\u8a08\u7b97\u76f8\u7576\u4f9d\u8cf4 \u8a5e\u983b\uff0c\u904e\u591a\u6216\u904e\u5c11\u90fd\u6703\u5f71\u97ff\u5176\u4fe1\u5fc3\u3002 \u4ee5\u672c\u5be6\u9a57\u70ba\u4f8b\uff0c\u8a5e\u983b\u751a\u4f4e\u7684\u6b63\u78ba\u8a5e\u5c0d\u61c9\u6709\u53ef\u80fd\u7121\u6cd5\u901a\u904e t-score \u7684\u7be9\u9078\u9580\u6abb\uff1b \u800c\u8a5e\u983b\u4e0d\u5920\u9ad8\u7684\u529f\u80fd\u8a5e\u5c0d\u61c9\u4e5f\u53ef\u80fd\u4ecd\u6709\u76f8\u7576\u9ad8\u7684 MI \u503c\u4ee5\u81f4\u65bc\u672a\u88ab\u904e\u6ffe\u3002 \u8a5e\u983b\u592a\u4f4e\u662f\u6240\u6709\u57fa\u65bc\u7d71\u8a08\u7684\u8a5e\u5c0d\u61c9\u90fd\u6703\u9762\u81e8\u7684\u56f0\u96e3\u554f\u984c\uff0c \u56e0\u70ba\u904e\u4f4e\u7684\u8a5e\u983b\u4e26\u6c92\u6709\u8fa6\u6cd5\u5206\u8fa8\u5176\u70ba \u5076\u7136\u6216\u662f\u6b63\u78ba\uff1b \u800c\u529f\u80fd\u8a5e\u7684\u90e8\u4efd\u53ef\u7528\u9810\u5148\u5efa\u7acb\u7684\u529f\u80fd\u8a5e\u5217\u8868\u4f86\u89e3\u6c7a\u3002 \u5728 recall \u65b9\u9762\uff0c\u7531\u65bc recall \u7684\u8a08\u7b97\u9700\u8981\u4ee5\u4eba\u5de5\u627e\u51fa\u96d9\u8a9e\u8a9e\u6599\u4e2d\u6240\u6709\u6b63\u78ba\u7684\u8a5e\u5c0d\u61c9\uff0c \u5728\u6b64\u5be6\u9a57 \u8cc7\u6599\u4e2d\uff0c\u5171\u6709 1192 \u500b\u76f8\u7570\u4e2d\u6587\u8a5e\u30011082 \u500b\u76f8\u7570\u82f1\u6587\u8a5e\uff0c \u57fa\u65bc\u6642\u9593\u53ca\u4eba\u529b\u7684\u95dc\u4fc2\u7121\u6cd5\u4ee5\u4eba\u5de5\u6a19 \u8a18\u6b64\u5be6\u9a57\u8cc7\u6599\u7684\u6240\u6709\u6b63\u78ba\u8a5e\u5c0d\u61c9\uff0c \u7136\u4ecd\u53ef\u77e5\u5728\u5206\u6bcd\u76f8\u540c\u7684\u689d\u4ef6\u4e0b\uff0c\u6211\u5011\u7684\u6f14\u7b97\u6cd5\u6709\u8f03\u9ad8\u7684 recall \u503c\u3002 \u7531\u65bc\u7ffb\u8b6f\u7684\u95dc\u4fc2\uff0c\u96d9\u8a9e\u5c0d\u8b6f\u7684\u7528\u8a5e\u53ef\u80fd\u76f8\u7576\u9748\u6d3b\uff0c \u4f8b\u5982\u540c\u4e00\u500b\u52d5\u8a5e\u537b\u5728\u4e0d\u540c\u4f4d\u7f6e\u7528\u4e0d\u540c\u7684 \u4e2d\u6587 (\u4f7f\u7528\u9017\u865f) 396 8.300 4.166 \u539f\u59cb\u6587\u7ae0\u4ee5 \u6a19\u9ede\u5206\u53e5 \u4e2d\u6587 104 31.615 18.074 \u82f1\u6587 165 23.666 12.524 \u4ee5\u672c\u6f14\u7b97\u6cd5 \u5206\u53e5 \u4e2d\u6587 265 12.384 9.683 \u82f1\u6587 265 14.709 9.172 \u5728\u5229\u7528 regular expression \u6216\u5176\u5b83\u65b9\u6cd5\u89e3\u6c7a\u82f1\u6587\u7e2e\u5beb\u9ede (\u5982\uff1aMr. \u6216 I.B.M.) \u7684\u554f\u984c\u4e4b\u5f8c\uff0c \u82f1 \u6587\u57fa\u672c\u4e0a\u53ef\u4ee5\u9760\u53e5\u865f\uff0c\u554f\u865f\uff0c\u9a5a\u5606\u865f\uff0c\u5206\u865f\u7576\u4f5c\u5206\u9694\u53e5\u5b50\u7684\u754c\u9650\u3002 \u4e2d\u6587\u7684\u53e5\u5b50\u7121\u6cd5\u50cf\u82f1\u6587\u4e00\u6a23 \u9760\u6a19\u9ede\u7b26\u865f\u4f86\u5224\u65b7\u3002 \u539f\u56e0\u662f\u9017\u9ede\u5728\u4e2d\u6587\u4f7f\u7528\u7684\u975e\u5e38\u7684\u9b06\u6563\uff0c \u9017\u865f\u548c\u53e5\u865f\u7684\u4f7f\u7528\u662f\u4f5c\u8005\u98a8\u683c\u7684\u554f \u984c\u800c\u975e\u6587\u6cd5\u7684\u554f\u984c\u3002 \u5982\u679c\u7528\u53e5\u865f\u3001\u554f\u865f\u3001\u9a5a\u5606\u865f\u3001\u5206\u865f\u4f86\u5206\u7684\u8a71\uff0c \u5f88\u591a\u662f\u6bd4\u53e5\u5b50\u66f4\u5927\u7684\u8a00\u8ac7\u55ae \u4f4d (discourse) \uff0c \u5982\u679c\u52a0\u4e0a\u9017\u865f\u7684\u8a71\u53c8\u6703\u9020\u6210\u8a31\u591a\u53ea\u662f\u8a5e\u7d44\u800c\u4e0d\u662f\u53e5\u5b50\u3002 \u9019\u5c31\u662f\u70ba\u4ec0\u9ebc\u7576\u6211\u5011\u7528 2.1 Warping \u7684\u6280\u8853\uff0c \u53ef\u8a08\u7b97\u5169\u5169\u8a5e\u5f59\u51fd\u5f0f\u5206\u4f48\u53d6\u6a23\u7684\u76f8\u4f3c\u7a0b\u5ea6\uff0c \u5f9e\u76f8\u4f3c\u7a0b\u5ea6\u7684\u9ad8\u4f4e\u53ef\u5f97\u5169\u8a5e\u5f59\u7684 \u76f8\u4f9d\u503c\u3002 \u4e0b\uff1a \u8a5e\u5f59\u7ffb\u8b6f\uff0c \u4ee5\u81f4\u65bc\u6240\u6709\u7684\u5c0d\u8b6f\u8a5e\u983b\u90fd\u4e0d\u9ad8\uff0c\u800c\u7121\u6cd5\u627e\u51fa\u6b63\u78ba\u8a5e\u5c0d\u61c9\u3002 \u7531\u65bc\u6b64\u539f\u56e0\uff0c\u8a31\u591a\u8a5e\u5f59\u7121 \u3002 \uff01 \uff1f \uff1b\u4f86\u5206\u5272\u53e5\u5b50\u6642\uff0c \u4e2d\u6587\u53e5\u6578\u6bd4\u82f1\u6587\u53e5\u5b50\u5c11\u5f88\u591a\uff0c \u800c\u4e2d\u6587\u52a0\u9017\u9ede\u4f5c\u70ba\u5206\u9694\u53e5\u5b50\u754c\u9650\u4e4b\u5f8c Wu et al. (2004) \u5247\u63d0\u51fa\u5229\u7528\u53e5\u9577\u548c\u6a19\u9ede\u7b26\u865f\u9032\u884c\u5c0f\u53e5\u5c0d\u61c9 (subsentential alignment)\uff0c \u52a0\u4e0a\u96d9 \u5728\u4e0a\u8ff0\u6f14\u7b97\u6cd5\u4e2d\uff0c\u5982\u679c\u52a0\u4e0a\u7279\u6b8a\u7684\u9650\u5236\u689d\u4ef6\uff0c\u5247\u53ef\u4f7f\u5207\u5206\u5340\u584a\u7684\u81ea\u7531\u5ea6\u964d\u4f4e\uff0c\u5f62\u6210\u7279\u5b9a\u7684 \u6cd5\u627e\u51fa\u6b63\u78ba\u8a5e\u5c0d\u61c9\uff0c \u800c\u6700\u5229\u65bc\u627e\u51fa\u7684\u5247\u662f\u6709\u56fa\u5b9a\u7ffb\u8b6f\u53ca\u9069\u7576\u8a5e\u983b\u7684\u5c08\u6709\u540d\u8a5e\u3002 \u53c8\u6bd4\u82f1\u6587\u53e5\u5b50\u591a\u5f88\u591a\u7684\u539f\u56e0\u3002 \u5f9e\u4ee5\u4e0a\u7684\u8a0e\u8ad6\uff0c\u6211\u5011\u53ef\u4ee5\u770b\u51fa\u7d93\u904e\u6211\u5011\u7684\u5206\u53e5\u6f14\u7b97\u6cd5\u6240\u5f97\u5230\u7684\u662f \u8a9e\u4e2d\u7684\u540c\u6e90\u8cc7\u8a0a (\u5982\u96d9\u8a9e\u4e2d\u76f8\u540c\u7684\u6578\u5b57\u90e8\u4efd)\uff0c \u4ee5\u9999\u6e2f\u7acb\u6cd5\u5c40\u6703\u8b70\u8a18\u9304\u70ba\u5be6\u9a57\u8cc7\u6599\uff0c\u53ef\u9054 98% \u7684 \u6b63\u78ba\u7387\u3002 Wu et al. (2004)\u6240\u7528\u7684\u82f1\u6f22\u5c0d\u8b6f\u8a9e\u6599\u70ba\u9999\u6e2f\u7acb\u6cd5\u5c40\u7684\u8b70\u6703\u7d00\u9304\uff0c \u5167\u5bb9\u5168\u662f\u8b70\u54e1\u8207\u5b98\u54e1 \u4e4b\u9593\u4e00\u554f\u4e00\u7b54\u7684\u7d00\u9304\u3002\u6b64\u985e\u8b70\u6703\u8a9e\u6599\u591a\u5c6c\u9010\u53e5\u7ffb\u8b6f\uff0c \u4e14\u5c11\u6709\u610f\u8b6f\u7684\u60c5\u5f62\uff0c\u7531\u65bc\u63a1\u4e00\u554f\u4e00\u7b54\u53ca\u9010 \u53e5\u7ffb\u8b6f\u5728\u53e5\u5c0d\u61c9\u53ca\u5c0f\u53e5\u5c0d\u61c9\u6bd4\u8f03\u5bb9\u6613\u3002 \u5982\u7528\u6587\u7ae0\u4e4b\u985e\u7684\u5c0d\u8b6f\u8a9e\u6599\u8a72\u6f14\u7b97\u6cd5\u52e2\u5fc5\u7121\u6cd5\u5f97\u5230\u5982\u6b64\u9ad8 \u5340\u584a\u3002 \u4f8b\u5982\u9650\u5236 start e|c \u7684\u524d\u4e00\u500b\u8a5e\u5fc5\u9808\u662f\u5206\u53e5\u7b26\u865f (\u5982\u53e5\u865f\u3001\u554f\u865f\u3001\u9a5a\u5606\u865f\u7b49)\uff0c end e|c \u5f8c\u4e00 \u6bb5\u843d\u6578 \u7e3d\u8a5e\u6578 \u76f8\u7570\u8a5e\u6578 \u6bd4\u53e5\u5b50\u9084\u8981\u5c0f\u7684\u5340\u584a\uff0c \u53ef\u8996\u70ba\u4e00\u7a2e\u5c0f\u53e5\u5c0d\u61c9\u7684\u7d50\u679c\u3002 \u4ee5\u4e0b\u8a66\u5217\u8209\u51fa\u524d 18 \u689d\u7531\u6211\u5011\u63d0\u51fa\u7684\u6f14\u7b97\u6cd5\u6240\u5f97\u5230\u7684\u8a5e\u5c0d\u61c9\u7d50\u679c\uff0c \u5176\u6b63\u78ba\u5c0d\u61c9\u8207\u5426\u65bc\u9644\u8a3b\u4e2d \u8a5e\u4e5f\u5fc5\u9808\u662f\u5206\u53e5\u7b26\u865f\uff0c\u5247\u6240\u5f97\u7684\u5340\u584a\u5c0d\u5c07\u6210\u70ba\u53e5\u5c0d\u61c9\u6216\u591a\u53e5\u5c0d\u61c9\u5f62\u5f0f\u3002 \u4ea6\u5373\u6b64\u6f14\u7b97\u6cd5\u70ba\u4e00\u8a5e\u5c0d \u61c9\u66a8\u53e5\u5c0d\u61c9\u4e4b\u6f14\u7b97\u6cd5\u3002 \u4e2d\u6587 59 3291 1192 \u82f1\u6587 59 3908 \u8aaa\u660e\uff0c\u82e5\u90e8\u4efd\u6b63\u78ba\u5247\u5728\u9644\u8a3b\u4e2d\u986f\u793a\u6b63\u78ba\u4e4b\u8a5e\u5c0d\u61c9\u3002 \u7531\u65bc\u6211\u5011\u7684\u6f14\u7b97\u6cd5\u4e26\u4e0d\u4fdd\u8b49\u6309\u9806\u5e8f\u5c0d\u61c9\uff0c \u56e0\u6b64\u8f38\u51fa\u7d50\u679c\u4e26\u4e0d\u6309\u539f\u59cb\u6587\u7ae0\u7684\u9806\u5e8f\u3002\u53e6\u5916\u57fa\u65bc 1082 \u6211\u5011\u6f14\u7b97\u6cd5\u7684\u7279\u6027\uff0c \u4e0d\u76f8\u9130\u7684\u5340\u584a\u6709\u88ab\u5408\u800c\u70ba\u4e00\u7684\u53ef\u80fd\u3002\u56e0\u70ba\u4e0a\u8ff0\u539f\u56e0\uff0c \u8981\u5c0d\u8f38\u51fa\u7684\u5c0d\u61c9\u5340 Figure 1: \u8f38\u51fa\u7d50\u679c\u6578\u8207\u6b63\u78ba\u6578\u95dc\u4fc2\u5716 \u7684\u6b63\u78ba\u7387\u3002 \u584a\u5206\u6790\u5176\u6b63\u78ba\u5206\u53e5\u7a0b\u5ea6\u6975\u70ba\u56f0\u96e3\u3002 \u56e0\u6b64\u6211\u5011\u63a1\u7528\u8f03\u7c21\u55ae\u7684\u4f30\u8a08\u65b9\u5f0f\uff0c \u4ee5\u4eba\u5de5\u6a19\u8a18 265 \u53e5\u4e2d\u5b8c Melamed (1998) \u7684 Competitive linking algorithm \u662f\u57fa\u65bc\u5df2\u6b63\u78ba\u53e5\u5c0d\u61c9\u7684\u8a5e\u5c0d\u61c9\u6f14\u7b97\u6cd5\u3002 \u5c0d\u65bc \u96d9\u8a9e\u7684\u5169\u5169\u8a5e\u5f59\uff0ccompetitive linking algorithm \u4f7f\u7528 LLR (Log-Likelihood-Ratio) \u4f86\u8a55\u4f30\u5176\u76f8\u4f9d\u7a0b \u5ea6\u3002 \u7576\u4e00\u96d9\u8a9e\u5c0d\u61c9\u53e5\u4e2d\u5169\u5169\u8a5e\u5f59\u7684\u76f8\u4f9d\u6b0a\u503c\u7686\u8a08\u7b97\u5b8c\u7562\uff0c \u5c07\u6240\u6709\u96d9\u8a9e\u8a5e\u5f59\u5c0d\u7531\u76f8\u4f9d\u6b0a\u503c\u5927\u81f3\u5c0f \u6392\u5e8f\uff0c\u4f9d\u5e8f\u53d6\u51fa\u96d9\u8a9e\u8a5e\u5f59\u5c0d\uff0c\u82e5\u5169\u8a5e\u5f59\u7686\u672a\u8207\u5176\u5b83\u8a5e\u5f59\u9023\u7d50\uff0c \u5247\u9023\u7d50\u6b64\u5169\u8a5e\u5f59\uff0c\u5426\u5247\u5ffd\u7565\u4e26\u8655 \u7406\u4e0b\u4e00\u8a5e\u5f59\u5c0d\uff0c\u76f4\u5230\u6240\u6709\u7684\u8a5e\u5f59\u5c0d\u7686\u5df2\u8655\u7406\u5b8c\u7562\u3002 3 \u6bb5\u843d\u5c0d\u9f4a\u5e73\u884c\u8a9e\u6599\u7684\u8a5e\u5c0d\u61c9\u66a8\u5c0f\u53e5\u5c0d\u61c9\u6f14\u7b97\u6cd5 3.1 \u6bb5\u843d\u5c0d\u9f4a\u5e73\u884c\u8a9e\u6599\u7684\u8a5e\u5c0d\u61c9\u6f14\u7b97\u6cd5 \u4e2d\u6587\u5206\u8a5e\u6b63\u78ba\u7387\u6700\u9ad8\u7684\u7cfb\u7d71\u4e4b\u4e00\u3002 6.363 table \u684c \u6b63\u78ba 5.948 Kuo \u90ed\u6167\u660e \u6b63\u78ba 5.626 hope \u5e0c\u671b \u6b63\u78ba 5.141 Jen-an \u4eba\u5b89 \u6b63\u78ba 4.948 each \u5929 each day \u6bcf\u5929 3.3 \u6e1b\u5c11\u53ef\u80fd\u5207\u5206\u65b9\u5f0f\u7684\u7e3d\u6578\u3002 \u7136\u800c\u8a08\u7b97\u6642\u9593\u4ecd\u7136\u76f8\u7576\u9577\uff0c\u56e0\u6b64\u96e3\u4ee5\u53d6\u5f97\u5ee3\u6cdb\u61c9\u7528\u3002 \u5728\u6b64\u6211\u5011\u63d0\u51fa \u82f1\u6587\u5206\u8a5e\u4ee5\u7a7a\u767d\u548c\u6a19\u9ede\u7b26\u865f\u70ba\u4e3b\uff0c\u642d\u914d\u5e38\u898b\u7e2e\u5beb\u8a5e\u4ee5\u6e1b\u5c11\u5206\u8a5e\u932f\u8aa4\u3002 \u8a08\u7b97\u82f1\u6587\u76f8\u7570\u8a5e\u6642\u5247 asso \u82f1\u6587\u8a5e \u4e2d\u6587\u8a5e \u9644\u8a3b \u5168\u6b63\u78ba\u3001\u90e8\u4efd\u6b63\u78ba\u53ca\u5b8c\u5168\u932f\u8aa4\u7684\u5c0f\u53e5\u5c0d\u61c9\uff0c \u5b8c\u5168\u6b63\u78ba\u8868\u793a\u8a72\u5c0d\u61c9\u662f\u6700\u5c0f\u53ef\u80fd\u7684\u5207\u5206\u65b9\u5f0f\uff0c\u4f8b</td></tr><tr><td>\u70ba\u4e00\u91cd\u8981\u7684\u7814\u7a76\u65b9\u5411\u3002 \u73fe\u6709\u7684\u96fb\u8166\u81ea\u52d5\u5efa\u7acb\u8a5e\u5c0d\u61c9\u7814\u7a76\u4e2d\uff0c\u6709\u8a31\u591a\u662f\u57fa\u65bc\u5df2\u6b63\u78ba\u53e5\u5c0d\u61c9\u7684\u7814\u7a76\uff0c \u4e26\u4e14\u53d6\u5f97\u4e0d\u932f\u7684 \u7814\u7a76\u6210\u679c\u3002 \u7136\u800c\u8981\u9054\u5230\u6b63\u78ba\u53e5\u5c0d\u61c9\u4e26\u4e0d\u5bb9\u6613\uff0c\u4ee5\u4eba\u5de5\u6a19\u793a\u8cbb\u6642\u8cbb\u529b\uff0c \u4e14\u5728\u73fe\u5be6\u74b0\u5883\u4e2d\uff0c\u4e26\u4e0d\u4fdd \u8b49\u6709\u6b63\u78ba\u53e5\u5c0d\u61c9\u3002\u800c\u4ee5\u6a5f\u5668\u81ea\u52d5\u53e5\u5c0d\u61c9\u7684\u6f14\u7b97\u6cd5\uff0c \u57fa\u65bc\u8a9e\u8a00\u7684\u7279\u6027\uff0c\u4e0d\u540c\u6027\u8cea\u7684\u6587\u7ae0\uff0c\u5176\u6b63\u78ba \u7387\u7684\u8b8a\u52d5\u975e\u5e38\u5927 (McEnery and Oakes, 1996)\uff0c \u8207\u5ee3\u6cdb\u61c9\u7528\u7684\u6c34\u6e96\u5c1a\u6709\u5dee\u8ddd\u3002 \u4ee5\u8a5e\u5c0d\u61c9\u7684\u89d2\u5ea6\u4f86 \u770b\uff0c\u6b63\u78ba\u7684\u8a5e\u5c0d\u61c9\u6709\u52a9\u65bc\u53e5\u5c0d\u61c9\uff1b\u800c\u5f9e\u53e5\u5c0d\u61c9\u4f86\u770b\uff0c \u6b63\u78ba\u7684\u53e5\u5c0d\u61c9\u5c0d\u8a5e\u5c0d\u61c9\u4e5f\u662f\u5341\u5206\u6b63\u9762\u7684\u52a9 \u76ca\u3002 \u53e5\u5c0d\u61c9\u548c\u8a5e\u5c0d\u61c9\u53ef\u8aaa\u662f\u96de\u751f\u86cb\u3001\u86cb\u751f\u96de\u7684\u554f\u984c\u3002 \u672c\u7814\u7a76\u4e3b\u8981\u63a2\u8a0e\u5982\u4f55\u5728\u540c\u4e00\u8a9e\u6599\u4e2d\u540c\u6642\u9032 \u884c\u53e5\u5c0d\u61c9\u53ca\u8a5e\u5c0d\u61c9\uff0c\u4e26\u85c9\u7531\u5f7c\u6b64\u63d0\u9ad8\u6b63\u78ba\u7387\u3002 \u6211\u5011\u9078\u64c7\u4f7f\u7528\u5df2\u6b63\u78ba\u6bb5\u843d\u5c0d\u61c9\u7684\u4e2d\u3001\u82f1\u8a9e\u6599\u5eab\uff0c \u7406\u7531\u5982\u4e0b\uff1a 1. \u57fa\u65bc\u7ffb\u8b6f\u7684\u7fd2\u6163\uff0c\u4ee5\u53e5\u70ba\u55ae\u4f4d\u4f86\u770b\uff0c\u5f80\u5f80\u6703\u6709\u589e\u6e1b\u7684\u60c5\u5f62\uff0c\u4f46\u82e5\u4ee5\u6bb5\u843d\u70ba\u55ae\u4f4d\u4f86\u770b\uff0c \u5247 \u8f03\u5c11\u6709\u589e\u6e1b\u7684\u60c5\u5f62\u3002\u56e0\u6b64\u4e00\u822c\u7684\u7ffb\u8b6f\u6587\u7ae0\u6216\u8005\u5df2\u6b63\u78ba\u53e5\u5c0d\u61c9\uff0c \u6216\u8005\u7d93\u904e\u6975\u5c11\u7684\u5de5\u4f5c\u5373\u53ef \u9054\u5230\u6b63\u78ba\u6bb5\u843d\u5c0d\u61c9\uff0c\u5c0d\u65bc\u73fe\u5be6\u7684\u61c9\u7528\u6709\u5f88\u5927\u7684\u5e6b\u52a9\u3002 2. \u7531\u65bc\u8a9e\u8a00\u7684\u7d50\u69cb\u95dc\u4fc2\uff0c\u6bb5\u843d\u8207\u6bb5\u843d\u9593\u5f80\u5f80\u6709\u5229\u65bc\u6a5f\u5668\u8655\u7406\u7684\u5206\u9694\u7b26\u865f\u5b58\u5728\uff0c \u56e0\u6b64\u6a5f\u5668\u81ea \u52d5\u5206\u6bb5\u53ef\u9054 100% \u6b63\u78ba\u3002\u56e0\u6b64\u4f7f\u7528\u6b63\u78ba\u6bb5\u843d\u5c0d\u61c9\u7684\u8a9e\u6599\u5eab\uff0c \u5728\u5206\u6bb5\u4e0a\u5e7e\u4e4e\u4e0d\u6703\u6709\u5931\u6557\u7684\u60c5 \u5f62\u767c\u751f\u3002 2.2 \u53e5\u5c0d\u61c9\u6f14\u7b97\u6cd5 \u96d9\u8a9e\uf906\u5c0d\u61c9\u7684\u7814\u7a76\u958b\u59cb\u65bc 90 \uf98e\u4ee3\u521d\u671f\u3002 Gale \u8207 Church (1991) \u53ca Brown \u7b49 (1991) \u89c0\u5bdf\u5230 \u9577\uf906\u7684\u7ffb\u8b6f\u5c0d\u61c9\uf906\u4e00\u822c\u800c\u8a00\u8f03\u9577\uff0c \u800c\u77ed\uf906\u7684\u7ffb\u8b6f\uf906\u901a\u5e38\u8f03\u77ed\u3002\u4ed6\u5011\uf9dd\u7528\uf906\u9577\u7684\u95dc\uf99a\u6027\u914d\u5408\u52d5\u614b \u898f\u5283\u6216 EM \u6f14\u7b97\u6cd5\u5f97\u5230 96% \u4ee5\u4e0a\u7684\u6b63\u78ba\uf961\u3002 Gale \u8207 Church (1991) \u53ca Brown \u7b49 (1991) \uf978\u8005\u6700\u5927 \u7684\u5dee\u5225\u662f\u524d\u8005\u900f\u904e\u4eba\u5de5\u5148\u5f97\u5230\u5148\u9a57\u6a5f\uf961 (prior probability) \u800c\u5f8c\u8005\uf9dd\u7528 EM \u6f14\u7b97\u6cd5\u5f97\u5230\u76f8\u95dc\u7684\uf96b \uf969\u3002Wu (1994) \u53ca Xu and Tan (1996) \u4ee5\uf906\u9577\u70ba\u4e3b\u7d50\u5408\u4e00\u500b\u5305\u542b\u65e5\u671f\u53ca\uf969\u5b57\u7b49\u8a0a\u606f\u5c0f\u7684\u8fad\u5178\u5f97\u5230 96% \u7684\u6b63\u78ba\uf961\u3002 \u4ee5\uf906\u9577\u70ba\u57fa\u790e\u7684\u7d71\u8a08\u65b9\u6cd5\u7684\u512a\u9ede\u662f\uf967\u9700\u8981\u8a9e\u8a00\u77e5\uf9fc\u53ca\u8fad\u5178\u5c31\u53ef\u4ee5\u904b\u4f5c\u3002 \u7f3a\u9ede\u662f \u5982\u679c\u8a9e\uf9be\u4e2d\u542b\u6709\u8c50\u5bcc\u7684\u591a\u5c0d\u591a\u7684\uf906\u5c0d\u61c9\u95dc\u4fc2\uff0c \u6216\u662f\u7ffb\u8b6f\u7684\u8a9e\uf9be\u4e2d\u6709\u589e\u6dfb\u6216\u522a\u6e1b\u7684\u73fe\u8c61\u767c\u751f\u5c31\u6703 \u9020\u6210\u6b63\u78ba\uf961\u5927\u5e45\u4e0b\ufa09\u3002 \u524d\u8ff0\u5e7e\u9805\u7814\u7a76\u7531\u65bc\u5927\u90fd\u63a1\u7528\u8b70\u6703\u7684\u7d00\uf93f\uff0c \uf9b5\u5982 Gale \u8207 Church (1991) \u53ca Brown \u7b49 (1991) \u7528\u52a0\u62ff\u5927\u570b\u6703 Hansard \u82f1\u6cd5\u5e73\ufa08\u8a9e\uf9be\uff0c Wu (1994) \u5247\uf9dd\u7528\u9999\u6e2f\uf9f7\u6cd5\u5c40\u8b70\u6703\u8cea\u8a62\u8207 \u7b54\u8a62\u7684\u4e2d\u82f1\u5e73\ufa08\u8a9e\uf9be\uff0c \u7531\u65bc\u662f\u53e3\u8a9e\u7d00\uf93f\u6240\u4ee5\uf906\u5b50\u8f03\u77ed\uff0c\u4e14\uf967\u5c11\u662f\u4e00\u5c0d\u4e00\u5c0d\u61c9\u3002 Gale \u8207 Church (1991) \u7d71\u8a08 Hansard \u8a9e\uf9be 80% \u4ee5\u4e0a\u662f\u4e00\u5c0d\u4e00\u7684\u5c0d\u61c9\u95dc\u4fc2\uff0c \u7f55\u6709\u591a\u5c0d\u591a\u7684\u5c0d\u61c9\u95dc\u4fc2\u6216\u589e\u6dfb\u6216\u522a \u6e1b\u7684\u60c5\u5f62\u767c\u751f\uff0c \u6240\u4ee5\u4ee5\uf906\u9577\u70ba\u4e3b\u7684\u7d71\u8a08\u65b9\u6cd5\u5f97\u5230\u5f88\u597d\u7684\u6548\u679c\u3002 \u4f46 McEnery and Oakes (1996) \u4ee5 Gale \u8207 Church (1991) \u7684\u65b9\u6cd5\u505a\u5be6\u9a57\u537b\u986f\u793a\u6b64\u7a2e\u6f14\u7b97\u6cd5\u7684\u6b63\u78ba\uf961\u5c0d\uf967\u540c\u7684\u6587\uf9d0\u8207\u8a9e\u8a00\u6703\u7522\u751f\u5f88\u5927 \u7684\u5dee\uf962\u3002 \uf9b5\u5982\u6ce2\uf91f\u6587\u82f1\u6587\u5e73\ufa08\u8a9e\uf9be\u7684\u6b63\u78ba\uf961\u56e0\u6587\uf9d0\uf967\u540c\u4ecb\u65bc\u65bc 100% \u8207 64.4%\uff0c \u800c\u4ed6\u5011\u6240\u5be6\u9a57 \u7684\u4e2d\u82f1\u65b0\u805e\u5e73\ufa08\u8a9e\uf9be\uf901\u4f4e\u65bc 55%\uff0c \u9019\u8b49\u660e\u55ae\u7d14\u4ee5\uf906\u9577\u95dc\uf99a\u6027\u986f\u7136\u7121\u6cd5\u5f97\u5230\u9ad8\u6b63\u78ba\uf961\u3002 \u53e6\u4e00\u500b\uf967\u9700\u8981\u8fad\u5178\u7684\u65b9\u6cd5\u662f Kay and R\u00f6scheisen (1993) \u4ee5\u8a5e\u5f59\u7684\u983b\uf961 (\u53bb\u9664\u4f4e\u983b\u7684\u8a5e\u53ca\u9ad8\u983b \u4e00\u500b\u52a0\u901f\u7684\u4f5c\u6cd5\u3002 5 \u5be6\u9a57\u7d50\u679c\u8207\u8a0e\u8ad6 4.877 volunteers \u7fa9\u5de5 \u6b63\u78ba \u5728 \u9375\u7684\u89d2\u8272\u5247\u662f\u6587\u7ae0\u7684\u5207\u5206\u3002 \u5c07\u6587\u7ae0\u5207\u5206\u6210\u82e5\u5e72\u5340\u584a\uff0c\u63d0\u4f9b\u4e86\u4e00\u500b\u5f37\u70c8\u7684\u5047\u8a2d\u53ca\u9650\u5236\uff0c \u5373\u8a72\u8a5e\u5f59 \u82e5\u6709\u8a5e\u5c0d\u61c9\uff0c\u5fc5\u7136\u51fa\u73fe\u65bc\u540c\u4e00\u5340\u584a\u4e2d\uff1b \u6b63\u78ba\u7684\u5207\u5206\u65b9\u5f0f\uff0c\u80fd\u4f7f\u8a5e\u5f59\u7684\u76f8\u4f9d\u7a0b\u5ea6\u63d0\u9ad8\uff0c\u53cd\u4e4b\u5247\u6703 \u964d\u4f4e\u3002 \u57fa\u65bc\u4e0a\u8ff0\u8aaa\u660e\uff0c\u6211\u5011\u8a2d\u8a08\u4e00\u6f14\u7b97\u6cd5\uff0c \u5728\u73fe\u6709\u7684\u5340\u584a\u4e2d\uff0c\u5c0b\u627e\u4e00\u5207\u5206\u65b9\u5f0f\uff0c\u53ef\u4f7f\u7e3d\u9ad4\u76f8\u4f9d \u503c\u63d0\u9ad8\u6700\u591a\uff0c \u4e0d\u65b7\u91cd\u8986\u6b64\u4e00\u904e\u7a0b\u76f4\u5230\u6240\u6709\u5207\u5206\u65b9\u5f0f\u90fd\u7121\u6cd5\u518d\u4f7f\u7e3d\u9ad4\u76f8\u4f9d\u503c\u63d0\u9ad8\u3002 \u4ee4 E = B e 1 B e 2 . . .\u3001C = B c 1 B c 2 . . .\uff0c \u5176\u4e2d B e|c 4.778 day \u5929 \u6b63\u78ba \u8003\u616e\u4e0a\u8ff0\u7684\u7406\u8ad6\u67b6\u69cb\uff0c\u5c0d\u65bc\u6bcf\u500b\u53ef\u80fd\u7684\u5207\u5206\u65b9\u5f0f\u90fd\u8981\u91cd\u65b0\u8a08\u7b97 ASSO \u503c\uff0c \u986f\u7136\u4ed8\u51fa\u592a\u5927\u7684 \u4ee3\u50f9\u3002 \u91cd\u65b0\u8a08\u7b97 ASSO \u503c\u7684\u7406\u7531\u5728\u65bc\u9019\u662f\u4e00\u500b\u8db3\u5920\u597d\u3001\u53ef\u4fe1\u8cf4\u7684\u8a55\u4f30\u65b9\u5f0f\uff0c \u53ef\u6709\u6548\u8a55\u4f30\u73fe\u884c 4.778 goal \u52aa\u529b \u932f\u8aa4 \u6211\u5011\u5be6\u4f5c\u4e86\u6211\u5011\u6240\u63d0\u51fa\u7684\u6f14\u7b97\u6cd5\uff0c\u4e26\u5be6\u4f5c K-vec \u6f14\u7b97\u6cd5\u4ee5\u9032\u884c\u6bd4\u8f03\u3002 \u5728 K-vec \u6f14\u7b97\u6cd5\u4e2d\uff0c \u7531\u65bc\u4f5c\u8005\u5efa\u8b70\u5207\u5206\u5340\u584a\u6578 K = \u221a total word number \u6703\u6709\u8f03\u597d\u7684\u7d50\u679c\uff0c \u800c\u5728\u6211\u5011\u7684\u5be6\u9a57\u8cc7\u6599\u4e2d\uff0c 4.725 fund \u7d93\u8cbb \u6b63\u78ba \u5206\u5272\u65b9\u5f0f\u7684\u512a\u52a3\u3002 \u56e0\u6b64\u52a0\u901f\u7684\u95dc\u9375\u5373\u5728\u65bc\u4f7f\u7528\u65b0\u7684\u8a55\u4f30\u65b9\u5f0f\uff0c \u65b0\u7684\u8a55\u4f30\u65b9\u5f0f\u9700\u6eff\u8db3\u4e0b\u5217\u689d\u4ef6\uff1a 1. \u548c ASSO \u76f8\u6bd4\u540c\u6a23\u53ef\u88ab\u4fe1\u8cf4\u3002 2. \u8a08\u7b97\u8907\u96dc\u5ea6\u8981\u4f4e\u3002 \u6211\u5011\u6240\u63d0\u51fa\u7684\u65b0\u8a55\u4f30\u65b9\u5f0f\u8aaa\u660e\u5982\u4e0b\uff1a \u5728\u4e00\u500b\u5340\u584a\u4e2d\uff0c\u6211\u5011\u53ef\u4ee5\u6307\u5b9a\u7684\u6a19\u9ede\u7b26\u865f (\u4f8b\u5982\u53e5\u865f\u3001\u554f\u865f\u7b49) \u5c07\u5340\u584a\u518d\u5207\u5206\u70ba\u8f03\u5c0f\u7684\u5340\u584a (\u53ef\u80fd\u5305\u542b\u4e00\u6216\u591a\u500b\u53e5\u5b50)\uff0c\u7a31\u70ba\u5b50\u5340\u584a\u3002\u4ee4 B e = S e 1 S e 2 . . . S e m \u548c B c = S c 1 S c 2 . . . S c n \uff0c\u5176\u4e2d B e \u548c B c \u70ba\u96d9\u8a9e\u8a9e\u6599\u4e2d\u5c0d\u61c9\u7684\u5176\u4e2d\u4e00\u5340\u584a\uff1b \u800c S c|e i \u8868\u793a\u5b50\u5340\u584a\u3002\u4ee4 W e|c i = {w e|c i,1 , w e|c i,2 , . . .} \u8868\u793a\u5728 S c|e i \u5b50\u5340\u584a\u4e2d\uff0c\u6240\u6709\u76f8\u7570\u8a5e\u5f59\u6240\u6210\u7684\u96c6\u5408\u3002 \u5b9a\u7fa9 4.626 Jen-an \u57fa\u91d1\u6703 The Jen-an Foundation \u4eba\u5b89\u57fa\u91d1\u6703 4.626 each \u6bcf \u6b63\u78ba 4.626 month \u6708 \u6b63\u78ba 4.533 even \u751a\u81f3 \u6b63\u78ba \u6bb5\u843d\u6578\u7684\u5e73\u65b9 (59 \u00d7 59 = 3481) K-vec 1.65 1.0 28 12 0.42 ours 1.65 1.0 79 32 0.40 4.488 but \u9577\u671f \u932f\u8aa4 4.404 welfare \u793e\u798f social welfare \u793e\u798f 4.247 social \u793e\u798f social welfare \u793e\u798f 4.141 elderly \u5931 elderly people \u4e09\u5931\u8001\u4eba 4.041 day \u6bcf each day \u6bcf\u5929 i value = max 1\u2264starte \u2264|E i | 1\u2264startc \u2264|C i | score(S e i , S c j ) = \u2211 e\u2208W e i \u2211 c\u2208W c Figure 2: \u8f38\u51fa\u7d50\u679c\u6578\u8207 precision \u95dc\u4fc2\u5716 asso(e, c) \u5728 \u5206 \u53e5 \u6f14 \u7b97 \u6cd5 \u65b9 \u9762 \uff0c \u6211 \u5011 \u7684 \u539f \u59cb \u8a9e \u6599 \u5171 \u6709 59 \u6bb5 \u843d \uff0c \u5728 \u5be6 \u4f5c \u4e2d \uff0c \u6211 \u5011 \u6307 \u5b9a \u9019 \u4e9b \u7b26 \u865f j starte \u2264ende \u2264|E i | startc \u2264endc\u2264|C i | . , ; ! ? \u3002 \uff0c \uff1b \uff01 \uff1f \u4f5c\u70ba\u5207\u5206\u5340\u584a\u7684\u6a19\u9ede\u7b26\u865f\u9650\u5236\u3002 \u7d93\u904e\u6211\u5011\u7684\u6f14\u7b97\u6cd5\uff0c\u6700\u5f8c\u6536\u6b5b\u6642\u5171\u8f38 \u5176\u4e2d asso(e, c) \u7684\u5b9a\u7fa9\u5982\u524d\u6240\u8ff0\u3002 \u5247\u85c9\u7531\u6c42\u51fa (max 1\u2264i\u2264|B e | 1\u2264j\u2264|B c | score(S e \u51fa 265 \u500b\u5c0d\u61c9\u5340\u584a\u3002 \u4e0b\u8868\u662f\u7d93\u7531\u6211\u5011\u7684\u6f14\u7b97\u6cd5\u7684\u5206\u53e5\u7d50\u679c\u8207\u539f\u59cb\u6587\u7ae0\u4ee5 . ; ! ? \u3002 \uff1b \uff01 \uff1f \u5206\u53e5\u7684 i , S c j ) \u6bd4\u8f03\u7d50\u679c\uff0c \u4e2d\u6587\u90e8\u4efd\u6211\u5011\u7528\u5169\u7d44\u6a19\u9ede\u7b26\u865f\u4f86\u5206\u53e5\uff0c\u5176\u4e2d\u4e00\u7d44\u5305\u542b\u9017\u865f\uff0c\u53e6\u4e00\u7d44\u4e0d\u5305\u542b\uff1a</td></tr></table>", |
| "html": null, |
| "type_str": "table" |
| } |
| } |
| } |
| } |
