| { | |
| "paper_id": "O05-1007", | |
| "header": { | |
| "generated_with": "S2ORC 1.0.0", | |
| "date_generated": "2023-01-19T07:58:31.285014Z" | |
| }, | |
| "title": "", | |
| "authors": [], | |
| "year": "", | |
| "venue": null, | |
| "identifiers": {}, | |
| "abstract": "", | |
| "pdf_parse": { | |
| "paper_id": "O05-1007", | |
| "_pdf_hash": "", | |
| "abstract": [], | |
| "body_text": [], | |
| "back_matter": [], | |
| "bib_entries": { | |
| "BIBREF0": { | |
| "ref_id": "b0", | |
| "title": "The Use of Topic Segmentation for Automatic Summarization", | |
| "authors": [ | |
| { | |
| "first": "R", | |
| "middle": [], | |
| "last": "Angheluta", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "R", | |
| "middle": [ | |
| "De" | |
| ], | |
| "last": "Busser", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "M.-F", | |
| "middle": [], | |
| "last": "Moens", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2002, | |
| "venue": "Proceedings of the ACL-2002 Post-Conference Workshop on Automatic Summarization", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "R. Angheluta and R. De Busser and M.-F. Moens, \"The Use of Topic Segmentation for Automatic Summarization,\" In Proceedings of the ACL-2002 Post-Conference Workshop on Automatic Summarization,2002.", | |
| "links": null | |
| }, | |
| "BIBREF1": { | |
| "ref_id": "b1", | |
| "title": "Using Lexical Chains for Text Summarization", | |
| "authors": [ | |
| { | |
| "first": "R", | |
| "middle": [], | |
| "last": "Barzilay", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "M", | |
| "middle": [], | |
| "last": "Elhadad", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1997, | |
| "venue": "ACL/EACL Workshop on Intelligent Scalable Text Summarization", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "10--17", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "R. Barzilay and M. Elhadad, \"Using Lexical Chains for Text Summarization,\" ACL/EACL Workshop on Intelligent Scalable Text Summarization, 1997 Page(s): 10 -17.", | |
| "links": null | |
| }, | |
| "BIBREF2": { | |
| "ref_id": "b2", | |
| "title": "Information fusion in the context of multi-document summarization", | |
| "authors": [ | |
| { | |
| "first": "R", | |
| "middle": [], | |
| "last": "Barzilay", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "K", | |
| "middle": [ | |
| "R" | |
| ], | |
| "last": "Mckeown", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "M", | |
| "middle": [], | |
| "last": "Elhadad", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1999, | |
| "venue": "Proceedings of the 37th Conference on Association for Computational Linguistics (ACL'99)", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "550--557", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Barzilay, R., McKeown, K. R., & Elhadad, M. (1999). Information fusion in the context of multi-document summarization. In Proceedings of the 37th Conference on Association for Computational Linguistics (ACL'99), College Park, Maryland, MD, USA (pp. 550-557).", | |
| "links": null | |
| }, | |
| "BIBREF3": { | |
| "ref_id": "b3", | |
| "title": "The use of MMR, diversity-based reranking for reordering documents and producing summaries", | |
| "authors": [ | |
| { | |
| "first": "J", | |
| "middle": [], | |
| "last": "Carbonell", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "J", | |
| "middle": [], | |
| "last": "Goldstein", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1999, | |
| "venue": "Proceedings of the 21st Annual International ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR'98)", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "335--336", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Carbonell, J., & Goldstein, J. (1999). The use of MMR, diversity-based reranking for reordering documents and producing summaries. In Proceedings of the 21st Annual International ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR'98), Melbourne, Australia (pp. 335-336).", | |
| "links": null | |
| }, | |
| "BIBREF4": { | |
| "ref_id": "b4", | |
| "title": "An Approach to Sentence-Selection-Based Text Summarization", | |
| "authors": [ | |
| { | |
| "first": "F", | |
| "middle": [], | |
| "last": "Chen", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "K", | |
| "middle": [], | |
| "last": "Han", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "G", | |
| "middle": [], | |
| "last": "Chen", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2002, | |
| "venue": "IEEE Region 10 Conference on Computers", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "489--493", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "F. Chen and K. Han and G. Chen, \"An Approach to Sentence-Selection-Based Text Summarization,\" IEEE Region 10 Conference on Computers, Communications, Control and Power Engineering, (TENCON '02) Volume1, Oct. 2002 Page(s):489-493.", | |
| "links": null | |
| }, | |
| "BIBREF5": { | |
| "ref_id": "b5", | |
| "title": "DUC 2003 (Document Understanding Conferences", | |
| "authors": [], | |
| "year": null, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "DUC 2003 (Document Understanding Conferences). Available at http://www-nlpir.nist.gov/projects/duc/guidelines/2003.html.", | |
| "links": null | |
| }, | |
| "BIBREF6": { | |
| "ref_id": "b6", | |
| "title": "Using argumentation to control lexical choice: a functional unification implementation", | |
| "authors": [ | |
| { | |
| "first": "M", | |
| "middle": [], | |
| "last": "Elhadad", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1993, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Elhadad, M. (1993). Using argumentation to control lexical choice: a functional unification implementation. Ph.D. Thesis, Department of Computer Science, Columbia University, New York, NY, USA.", | |
| "links": null | |
| }, | |
| "BIBREF7": { | |
| "ref_id": "b7", | |
| "title": "Detecting similarity by applying learning over indicators", | |
| "authors": [ | |
| { | |
| "first": "E", | |
| "middle": [], | |
| "last": "Eskin", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "J", | |
| "middle": [], | |
| "last": "Klavans", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "V", | |
| "middle": [], | |
| "last": "Hatzivassiloglou", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1999, | |
| "venue": "Proceedings of the 37th Conference on Association for Computational Linguistics (ACL'99)", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Eskin, E., Klavans, J., & Hatzivassiloglou, V. (1999). Detecting similarity by applying learning over indicators. In Proceedings of the 37th Conference on Association for Computational Linguistics (ACL'99), College Park, Maryland, MD, USA.", | |
| "links": null | |
| }, | |
| "BIBREF9": { | |
| "ref_id": "b9", | |
| "title": "Multi-document summarization by sentence extraction", | |
| "authors": [ | |
| { | |
| "first": "J", | |
| "middle": [], | |
| "last": "Goldstein", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "V", | |
| "middle": [], | |
| "last": "Mittal", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "J", | |
| "middle": [], | |
| "last": "Carbonell", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "M", | |
| "middle": [], | |
| "last": "Kantrowitz", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2000, | |
| "venue": "Proceedings the ANLP/NAACL Workshop on Automatic Summarization", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "40--48", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Goldstein, J., Mittal, V., Carbonell, J., & Kantrowitz, M. (2000). Multi-document summarization by sentence extraction. In Proceedings the ANLP/NAACL Workshop on Automatic Summarization, Seattle, WA (pp. 40-48).", | |
| "links": null | |
| }, | |
| "BIBREF10": { | |
| "ref_id": "b10", | |
| "title": "Data Mining Concepts and Techniques", | |
| "authors": [ | |
| { | |
| "first": "J", | |
| "middle": [], | |
| "last": "Han", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "M", | |
| "middle": [], | |
| "last": "Kamber", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2000, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "J. Han and M. Kamber, Data Mining Concepts and Techniques, Morgan Kaufmann, 2000.", | |
| "links": null | |
| }, | |
| "BIBREF12": { | |
| "ref_id": "b12", | |
| "title": "FIDS: an intelligent financial Web news articles digest system", | |
| "authors": [ | |
| { | |
| "first": "W", | |
| "middle": [], | |
| "last": "Lam", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "K", | |
| "middle": [ | |
| "S" | |
| ], | |
| "last": "Ho", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2001, | |
| "venue": "IEEE Transactions on Systems, Man and Cybernetics, Part A", | |
| "volume": "31", | |
| "issue": "6", | |
| "pages": "753--762", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "W. Lam and K. S. Ho, \"FIDS: an intelligent financial Web news articles digest system,\" IEEE Transactions on Systems, Man and Cybernetics, Part A, Volume 31, Issue 6, Nov. 2001 Page(s):753 -762.", | |
| "links": null | |
| }, | |
| "BIBREF13": { | |
| "ref_id": "b13", | |
| "title": "A Fuzzy Ontology and Its Application to News Summarization", | |
| "authors": [ | |
| { | |
| "first": "C", | |
| "middle": [ | |
| "S" | |
| ], | |
| "last": "Lee", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Z", | |
| "middle": [ | |
| "W" | |
| ], | |
| "last": "Jian", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "L", | |
| "middle": [ | |
| "K" | |
| ], | |
| "last": "Huang", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2005, | |
| "venue": "IEEE Transactions on Systems, Man, and Cybernetics-Part B: Cybernetics : Accepted", | |
| "volume": "99", | |
| "issue": "", | |
| "pages": "859--880", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "C. S. Lee, Z. W. Jian and L. K. Huang, \"A Fuzzy Ontology and Its Application to News Summarization,\" IEEE Transactions on Systems, Man, and Cybernetics-Part B: Cybernetics : Accepted for future publication Volume PP, Issue 99, 2005 Page(s):859 -880.", | |
| "links": null | |
| }, | |
| "BIBREF14": { | |
| "ref_id": "b14", | |
| "title": "Summarizing similarities and differences among related documents", | |
| "authors": [ | |
| { | |
| "first": "I", | |
| "middle": [], | |
| "last": "Mani", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "E", | |
| "middle": [], | |
| "last": "Bloedorn", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1999, | |
| "venue": "Information Retrieval", | |
| "volume": "1", | |
| "issue": "1-2", | |
| "pages": "35--67", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Mani, I., & Bloedorn, E. (1999). Summarizing similarities and differences among related documents. Information Retrieval, 1(1-2), 35-67.", | |
| "links": null | |
| }, | |
| "BIBREF15": { | |
| "ref_id": "b15", | |
| "title": "Using sentence-selection heuristics to rank text segment in TXTRACTOR", | |
| "authors": [ | |
| { | |
| "first": "D", | |
| "middle": [], | |
| "last": "Mcdonald", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "H", | |
| "middle": [ | |
| "C" | |
| ], | |
| "last": "Chen", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2002, | |
| "venue": "Proceedings of the second ACM/IEEE-CS joint conference on Digital libraries", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "28--35", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "D. McDonald and H.C. Chen, \"Using sentence-selection heuristics to rank text segment in TXTRACTOR,\" Proceedings of the second ACM/IEEE-CS joint conference on Digital libraries, Portland, Oregon, USA, 2002 Page(s): 28 -35.", | |
| "links": null | |
| }, | |
| "BIBREF16": { | |
| "ref_id": "b16", | |
| "title": "Towards multidocument summarization by reformulation: progress and prospects", | |
| "authors": [ | |
| { | |
| "first": "K", | |
| "middle": [ | |
| "R" | |
| ], | |
| "last": "Mckeown", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "J", | |
| "middle": [ | |
| "L" | |
| ], | |
| "last": "Klavans", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "V", | |
| "middle": [], | |
| "last": "Hatzivassiloglou", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "R", | |
| "middle": [], | |
| "last": "Barzilay", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "E", | |
| "middle": [], | |
| "last": "Eskin", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1999, | |
| "venue": "Proceedings of the 16th National Conference on Artificial Intelligence (AAAI'99)", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "453--460", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "McKeown, K. R., Klavans, J. L., Hatzivassiloglou, V., Barzilay, R., & Eskin, E. (1999). Towards multidocument summarization by reformulation: progress and prospects. In Proceedings of the 16th National Conference on Artificial Intelligence (AAAI'99), Orlando, FA, USA (pp. 453-460).", | |
| "links": null | |
| }, | |
| "BIBREF17": { | |
| "ref_id": "b17", | |
| "title": "Text Mining with Information Extraction", | |
| "authors": [ | |
| { | |
| "first": "U", | |
| "middle": [ | |
| "Y" | |
| ], | |
| "last": "Nahm", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "R", | |
| "middle": [ | |
| "J" | |
| ], | |
| "last": "Mooney", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2002, | |
| "venue": "Proceedings of the AAAI-2002 Spring Symposium on Mining Answers from Texts and Knowledge Bases", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "U. Y. Nahm and R. J. Mooney, \"Text Mining with Information Extraction,\" In Proceedings of the AAAI-2002 Spring Symposium on Mining Answers from Texts and Knowledge Bases, 2002.", | |
| "links": null | |
| }, | |
| "BIBREF18": { | |
| "ref_id": "b18", | |
| "title": "NLPprocess-Text Analysis Toolkit", | |
| "authors": [], | |
| "year": null, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "NLPprocess-Text Analysis Toolkit. Available at http://www.infogistics.com/textanalysis.html.", | |
| "links": null | |
| }, | |
| "BIBREF19": { | |
| "ref_id": "b19", | |
| "title": "Revision-based generation of natural language summaries providing historical background: corpus-based analysis, design, implementation and evaluation", | |
| "authors": [ | |
| { | |
| "first": "J", | |
| "middle": [], | |
| "last": "Robin", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1994, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Robin, J. (1994). Revision-based generation of natural language summaries providing historical background: corpus-based analysis, design, implementation and evaluation. Ph.D. Thesis, Department of Computer Science, Columbia University, New York, NY, USA.", | |
| "links": null | |
| }, | |
| "BIBREF20": { | |
| "ref_id": "b20", | |
| "title": "A Non-Linear Topic Detection Method for Text Summarization Using Wordnet", | |
| "authors": [ | |
| { | |
| "first": "C", | |
| "middle": [ | |
| "N" | |
| ], | |
| "last": "Silla", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "C", | |
| "middle": [ | |
| "A A" | |
| ], | |
| "last": "Kaestner", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "A", | |
| "middle": [ | |
| "A" | |
| ], | |
| "last": "Freitas", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2003, | |
| "venue": "Workshop of Technology Information Language Human (TIL'2003)", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "C. N. Silla Jr. and C. A. A. Kaestner and A. A.Freitas, \"A Non-Linear Topic Detection Method for Text Summarization Using Wordnet,\" Workshop of Technology Information Language Human (TIL'2003), 2003.", | |
| "links": null | |
| }, | |
| "BIBREF21": { | |
| "ref_id": "b21", | |
| "title": "The Porter Stemming Algorithm", | |
| "authors": [], | |
| "year": null, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "The Porter Stemming Algorithm. Available at http://tartarus.org/~martin/PorterStemmer.", | |
| "links": null | |
| }, | |
| "BIBREF22": { | |
| "ref_id": "b22", | |
| "title": "Event-Centric Summary Generation", | |
| "authors": [ | |
| { | |
| "first": "L", | |
| "middle": [], | |
| "last": "Vanderwende", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "M", | |
| "middle": [], | |
| "last": "Banko", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "A", | |
| "middle": [], | |
| "last": "Menezes", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2004, | |
| "venue": "Document Understanding Conference at HLT-NAACL", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "L. Vanderwende and M. Banko and A. Menezes, \"Event-Centric Summary Generation,\" In Document Understanding Conference at HLT-NAACL, Boston, MA, 2004.", | |
| "links": null | |
| }, | |
| "BIBREF23": { | |
| "ref_id": "b23", | |
| "title": "Simplex NPs clustered by head: a method for identifying significant topics in a document", | |
| "authors": [ | |
| { | |
| "first": "N", | |
| "middle": [], | |
| "last": "Wacholder", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1998, | |
| "venue": "Proceedings of Workshop on the Computational Treatment of Nominals, COLING-ACL", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "70--79", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Wacholder, N. (1998). Simplex NPs clustered by head: a method for identifying significant topics in a document. In Proceedings of Workshop on the Computational Treatment of Nominals, COLING-ACL, Montreal, Canada (pp. 70-79).", | |
| "links": null | |
| } | |
| }, | |
| "ref_entries": { | |
| "TABREF0": { | |
| "num": null, | |
| "html": null, | |
| "type_str": "table", | |
| "text": "Document Understanding Conference) \u6240\u63d0\u4f9b\u7684\u65b0\u805e\u4e8b\u4ef6\u9032\ufa08\u8a55\u4f30\u3002\u8a55\u4f30\u7d50\u679c\u65bc ROUGE-1 \u6307\u6a19\u6bd4\u5e73\u5747\u6210\u7e3e\u9084\u597d\uff0c\u65bc ROUGE-L \u6307\u6a19\u63a5\u8fd1\u6700\u597d\u7684\u7d50\u679c\u3002 (P ij ,Q,C ij )\u8a08\u7b97 P ij \u8207 Q \u7684\u76f8\u4f3c\ufa01\uff0c\u540c\u6642\u8861\uf97e\u8207\u6bb5\uf918\u6240\u5728\u7684\u6587\u4ef6\u7fa4\u7684\u76f8\u95dc\ufa01\uff1b Sim 2 (P ij ,P nm ,C,S)\u8a08\u7b97 P ij \u8207 P nm \u7684\u76f8\u4f3c\ufa01\uff0c\u5176\u4e2d P nm \u70ba\u4e00\u4ee5\u6311\u9078\u51fa\u4e4b\u6bb5\uf918\u3002 \u500b\u5b57\u4f5c\u70ba\u63cf\u8ff0\u5b57\u5f59\u96c6\u5408\uff0c\u540c\u6642\u9650\u5236\u6311\u9078\u7684\u7bc4\u570d\u70ba\u4e00\u500b\u5b8c\u6574\u7684\u8a9e\uf906\u3002\u5728 N \u8207 M \u7684 \u8a2d\u5b9a\u4e0a\uff0c\u7531\u65bc\u5728[5]\u4e2d\u63d0\u53ca\u4eba\uf9d0\u7684\u77ed\u66ab\u8a18\u61b6\u901a\u5e38\u70ba 7\u00b12 \u500b\u5b57\u8a5e\uff0c\u56e0\u6b64\uff0c\u5be6\u4f5c\u4e0a\u8a2d\u5b9a N \u53ca M \u5404\u70ba 5\u3002\u53d6\u6700\u5c0f\u503c\u6700\u4e3b\u8981\u662f\u8981\u8b93\u524d\u5f8c\u6587\u7684\u6db5\u84cb\u7bc4\u570d\u5c0f\u4e00\u9ede\uff0c\u4f7f\u76f8\u9130\u7684\uf967\u540c\u6982\uf9a3\u5728\u63cf\u8ff0\u7684\u5167\u5bb9\uf967\u81f4\u65bc\u6709\u904e\u591a \u91cd\u8907\uff0c\u4ee5\u514d\u5f71\u97ff\u5230\u6982\uf9a3\u5206\u7fa4\u7684\u7d50\u679c\u3002 \u8868 1 \u70ba\u4e0b\uf9b5\u4e2d\u300cthe U.S. Embassy\u300d\u7684\u63cf\u8ff0\u6cd5\uff0c\u4ee5 several locations, Bonn, receiving, word, a terrorist threat, the U.S. Embassy, no evidence, a planned attack, found, officials, Wednesday \u4f5c\u70ba\uf96a\u5f15 \u8a5e(Indexing Description)\u3002\u6b64\u63cf\u8ff0\u6982\uf9a3\u7684\u65b9\u5f0f\u662f\u5e0c\u671b\uf9dd\u7528\u524d\u5f8c\u6587\u7684\u95dc\u4fc2\uff0c\u8b93\u6982\uf9a3\u7684\u8a9e\u610f\uf901\u52a0\u660e\u986f\uff0c \u4ee5\u65b9\uf965\u9032\ufa08\u5206\u7fa4\u7684\u6642\u5019\u80fd\uf901\u7cbe\u78ba\u8a08\u7b97\uf978\uf978\u6982\uf9a3\u7684\u76f8\u4f3c\ufa01\u3002 BONN, Germany (AP) _", | |
| "content": "<table><tr><td>(Semantic Network)\uff0c\u63cf\u8ff0\u6982\uf9a3\u8a5e\u5f59\uff1b2) \u4f7f\u7528\u5206\u7fa4\u6cd5(\u672c\u6587\u63a1\u7528 K-Means [11])\u5c0d\u6982\uf9a3\u8a5e\u5206\u7fa4\uff0c\u4ee5\u8403 \u53d6\uf901\u7cbe\u78ba\u7684\u6982\uf9a3\uff0c\u540c\u6642\u89e3\u6c7a\u8a9e\u610f\u6b67\uf962\u7684\u554f\u984c\uff1b3) \u6839\u64da\u6982\uf9a3\u5206\u7fa4\u7d50\u679c\uff0c\uf9dd\u7528\u8a9e\uf906\u8cc7\u8a0a\uf97e\u3001\u8a9e\uf906\u4f4d )] , , , ( max ) 1 ( ) , , ( 1 [ max 2 / S C P P Sim C Q P Sim Arg MD MMR nm ij S P ij ij S R P def ij ij \u2208 \u2208 \u2212 \u2212 = \u2212 \u03bb \u03bb \u6790\uff0c\u767c\u73fe\u300cthe U.S. Embassy\u300d\u8207\u4e0a\u8ff0\u4e94\u500b\u5b57\u4e26\u6c92\u6709\u76f8\u95dc\uff0c\u56e0\u6b64\u5728\u8868 2 \u4e2d\uf965\u53bb\u9664\u6b64\u4e94\u500b\u5b57\u5f59\u3002\u7531 \u6b64\u53ef\u770b\u51fa\u52a0\u5165\u8a9e\u610f\u7db2\uf937\u5f8c\uff0c\u53ef\u6d88\u9664\u591a\u9918\u7684\u5b57\u8a5e\uff0c\u4f7f\u5f97\u63cf\u8ff0\u6982\uf9a3\u7684\u5b57\u8a5e\uf901\u7cbe\u78ba\u3002 \uf9a3\u7fa4\u7684\u76f8\u4f3c\ufa01\u3002 \uf906\u6240\u5305\u542b\u6982\uf9a3\u8207\u6982\uf9a3\u7fa4\u7684\u76f8\u4f3c\ufa01\u3002\u672c\u6587\u91dd\u5c0d\u6b64\uf978\u7a2e\u5c0d\u61c9\u65b9\u5f0f\uff0c\u63d0\u51fa\uf967\u540c\u8a08\u7b97\u76f8\u4f3c\ufa01\u7684\u65b9\u6cd5\u3002 ( ) (S S sim location \u03bb \u03b8 + ) ) (C ) (S ) (C _ distance tfidf length \u03b3 \u03b2 \u03b1 + + + = weight sentence</td></tr><tr><td>\u6458\u8981. \u65b0\u805e\u4e8b\u4ef6\u81ea\u52d5\u6458\u8981\u4e43\u91dd\u5c0d\u6558\u8ff0\u76f8\u4f3c\u4e8b\u4ef6\u7684\u591a\u7bc7\u65b0\u805e\u6587\u7ae0\u7de8\u88fd\u6458\u8981\u5167\u5bb9\uff0c\u5176\u76ee\u7684\u70ba\u5e6b\u52a9\uf95a\u8005 \u904e\uf984\u8cc7\u8a0a\u4e26\u5feb\u901f\u77ad\u89e3\u4e8b\u4ef6\u7684\uf92d\uf9c4\u53bb\u8108\uff0c\u4ee5\u7bc0\uf96d\u95b1\uf95a\u5927\uf97e\u65b0\u805e\u6587\u4ef6\u7684\u6642\u9593\uff0c\u4e3b\u8981\u7684\u7814\u7a76\u8b70\u984c\u70ba\u5075\u6e2c \uf967\u540c\u65b0\u805e\u6587\u7ae0\u4e2d\u76f8\u4f3c\u53ca\u76f8\uf962\u7684\u5167\u5bb9\uff0c\u4ee5\u9054\u5230\u904e\uf984\u91cd\u8907\u8cc7\u8a0a\u7684\u76ee\u7684\u3002\u672c\uf941\u6587\u4ee5\u6982\uf9a3\u5206\u7fa4(Concept Clustering)\u70ba\u57fa\u790e\uff0c\u5075\u6e2c\u65b0\u805e\u4e8b\u4ef6\u6240\u8981\u8868\u9054\u7684\u8a9e\u610f\uff0c\u9032\u800c\u6311\u9078\u6db5\u84cb\u8c50\u5bcc\u8a9e\u610f\u7684\u8a9e\uf906\u70ba\u6458\u8981\u3002\u904e\u7a0b \u70ba\uff1a1) \uf9dd\u7528\u524d\u5f8c\u6587\u95dc\u4fc2(Context)\u53ca\u8a9e\u610f\u7db2\uf937(Semantic Network)\uff0c\u63cf\u8ff0\u6982\uf9a3\u8a5e\u5f59\uff1b2) \u4f7f\u7528 K-Means \u5c0d\u6982\uf9a3\u8a5e\u5206\u7fa4\uff0c\u671f\u8403\u53d6\uf901\u7cbe\u78ba\u7684\u6982\uf9a3\uff0c\u540c\u6642\u89e3\u6c7a\u8a9e\u610f\u6b67\uf962\u7684\u554f\u984c\uff1b3) \u6839\u64da\u6982\uf9a3\u5206\u7fa4\u7d50\u679c\uff0c\u4e26\uf9dd \u7528\u8a9e\uf906\u8cc7\u8a0a\uf97e\u3001\u8a9e\uf906\u4f4d\u7f6e\u53ca\u8a9e\uf906\u6982\uf9a3\u7b49\u7279\u5fb5\uff0c\u8a08\u7b97\u6bcf\u500b\u8a9e\uf906\u4e4b\u91cd\u8981\u6027\uff0c\u6700\u5f8c\u6311\u9078\u91cd\u8981\u6027\u9ad8\u7684\u8a9e\uf906 \u4f5c\u70ba\u6458\u8981\u5167\u5bb9\u3002\u5be6\u9a57\u4e2d\u4f7f\u7528 DUC 2003 (1. \u524d\u8a00 \u8fd1\uf98e\uf92d\uff0c\u7531\u65bc\u96fb\u8166\u79d1\u6280\u7684\u8fc5\u901f\u767c\u5c55\u53ca\u7db2\u969b\u7db2\uf937\u7684\u63a8\u6ce2\u52a9\u703e\uff0c\u8cc7\u8a0a\uf9d3\u7e8c\u88ab\uf969\u4f4d\u5316\uff0c\u4ee5\uf9dd\u65bc\u7db2 \uf937\uf9ca\u50b3\u3002\uf969\u4f4d\u5316\u7684\u767c\u5c55\u4ea6\u4f7f\u5f97\u8cc7\u8a0a\uf97e\u5927\u5e45\u589e\u52a0\uff0c\u4f7f\u7528\u8005\u5728\u7372\u53d6\u8cc7\u8a0a\u4e0a\uf967\u518d\u53d7\u9650\u65bc\u5c11\uf969\u7684\uf9ca\u901a\u7ba1 \u9053\uff0c\u53cd\u800c\u53ef\u4ee5\u8f15\uf9e0\u53d6\u5f97\u5927\uf97e\u8cc7\uf9be\u3002\u5728\u9019\u7a2e\u73fe\u8c61\u4e0b\uff0c\u56f0\u96e3\u7684\u53cd\u800c\u662f\u5982\u4f55\u904e\uf984\u6389\uf967\u9700\u8981\u53ca\u91cd\u8907\u7684\u8cc7\u8a0a\uff0c \u4f7f\u5f97\u4f7f\u7528\u8005\u53ef\u4ee5\u5feb\u901f\u627e\u5230\u771f\u6b63\u6240\u9700\u7684\u8cc7\u8a0a\u3002\u70ba\u89e3\u6c7a\u524d\u8ff0\u56f0\u96e3\uff0c\u53ef\uf9dd\u7528\u6458\u8981\u7cfb\u7d71\u7684\u7cbe\u7c21\u6027\u53ca\u53bb\u91cd\u8907 \u6027\uff0c\u6e1b\u5c11\u4f7f\u7528\u8005\u95b1\uf95a\u6642\u9593\uff0c\u5e6b\u52a9\u4f7f\u7528\u8005\u65bc\u77ed\u6642\u9593\u5224\u65b7\u53ca\u53d6\u5f97\u91cd\u8981\u8cc7\u8a0a\u3002 \u6587\u4ef6\u6458\u8981\u6280\u8853\u7684\u4f5c\u6cd5\u5927\u81f4\u53ef\u5206\u70ba\uf978\uf9d0\u3002\u7b2c\u4e00\uf9d0\u4f7f\u7528\u5c08\u696d\uf9b4\u57df\u77e5\uf9fc(Domain Knowledge)\u5206\u6790\u6587 \u7ae0\u4e2d\u7684\u4eba\u3001\u4e8b\u3001\u6642\u3001\u5730\u3001\u7269\u7b49\u8981\u7d20\uff0c\u7b2c\u4e8c\uf9d0\u5247\u662f\u4ee5\u7d71\u8a08\u5206\u6790(Statistical Analysis)\u65b9\u6cd5\u76f4\u63a5\u5f9e\u539f\u6587 \u5224\u65b7\u8a9e\uf906\u91cd\u8981\u6027\u3002\u4f7f\u7528\u5c08\u696d\uf9b4\u57df\u77e5\uf9fc\uf92d\u9054\u6210\u6458\u8981\u7cfb\u7d71\u53ef\u6709\u6548\u62bd\u53d6\u51fa\u6587\u4ef6\u5167\u7684\u4e3b\u984c\uff0c\u4f46\u662f\u9700\u8981\u4e8b\u524d \u7531\uf9b4\u57df\u5c08\u5bb6\u4ecb\u5165\u5efa\uf9f7\uf9b4\u57df\u77e5\uf9fc\uff0c\u5305\u62ec\u8a9e\u8a00\u77e5\uf9fc\u3001\u6587\u4ef6\u4e3b\u984c\u80cc\u666f\u77e5\uf9fc\u7b49\uff0c\u800c\u81ea\u52d5\u5316\u5efa\uf9f7\uf9b4\u57df\u77e5\uf9fc\u7684 \u65b9\u6cd5\u76ee\u524d\u9084\u5f88\u96e3\u7a81\u7834\u3002\u672c\uf941\u6587\u4ee5\u8cc7\u8a0a\u64f7\u53d6(Information Retrieval)\u6280\u8853\u70ba\u57fa\u790e\uff0c\u5c0e\u5165\u8a9e\u610f\u7db2\uf937 (Semantic Network)\uff0c\u540c\u6642\u6539\uf97c\u539f\u6587\u62bd\u53d6\u6458\u8981\u8a9e\uf906\u7684\u65b9\u5f0f\uff0c\u63d0\u51fa\u4e00\u5957\u63cf\u8ff0\u6982\uf9a3(Concept)\u4e14\u80fd\u5206\u8fa8 \u8a9e\u610f(Semantics)\u7684\u6982\uf9a3\u5075\u6e2c\u65b9\u6cd5\u3002\uf9dd\u7528\uf967\u540c\u7684\u6982\uf9a3\uff0c\uf965\u53ef\u627e\u51fa\u65b0\u805e\u4e8b\u4ef6\u4e2d\u5177\u8c50\u5bcc\u6982\uf9a3\u53ca\u8a9e\u610f\u7684\u8a9e \u7f6e\u53ca\u8a9e\uf906\u6982\uf9a3\u7b49\u7279\u5fb5\uff0c\u8a08\u7b97\u6bcf\u500b\u8a9e\uf906\u7684\u91cd\u8981\u6027\uff0c\u6700\u5f8c\u6311\u9078\u91cd\u8981\u6027\u9ad8\u7684\u8a9e\uf906\u4f5c\u70ba\u6458\u8981\u5167\u5bb9\u3002 \u672c\u6587\u5171\u5206\u6210\uf9d1\u7bc0\u3002\u7b2c\u4e8c\u7bc0\u4ecb\u7d39\u8207\u591a\u6587\u4ef6\u81ea\u52d5\u6458\u8981\u76f8\u95dc\u7684\u7814\u7a76\uff1b\u7b2c\u4e09\u7bc0\u4ecb\u7d39\u7d50\u5408\u524d\u5f8c\u6587\u8207\u8a9e \u610f\u7db2\uf937\u7684\u6982\uf9a3\u63cf\u8ff0\u6cd5\uff0c\u4e26\uf96f\u660e\u5982\u4f55\u9032\ufa08\u5206\u7fa4\u5075\u6e2c\u53ca\u62bd\u53d6\u6982\uf9a3\uff1b\u7b2c\u56db\u7bc0\u91dd\u5c0d\u5148\u524d\u9078\u51fa\u7684\u6982\uf9a3\u627e\u51fa\u5c0d \u61c9\u7684\u8a9e\uf906\uff0c\u4e26\u4ee5\uf969\u500b\u7279\u5fb5\u503c\u9032\ufa08\u8a9e\uf906\u7684\u6b0a\u91cd\u5206\u6790\uff1b\u7b2c\u4e94\u7bc0\uf96f\u660e\u5be6\u9a57\u7d50\u679c\u7684\u5206\u6790\u8a0e\uf941\uff0c\u4ee5\u9a57\u8b49\u672c\u6587 \u6240\u63d0\u65b9\u6cd5\u7684\u53ef\ufa08\u6027\uff1b\u6700\u5f8c\u4e00\u7bc0\u662f\u7d50\uf941\u8207\u672a\uf92d\u53ef\u7e7c\u7e8c\u767c\u5c55\u7684\u65b9\u5411\u3002 2. \u76f8\u95dc\u7814\u7a76\u5de5\u4f5c \u672c\u7bc0\u4ecb\u7d39\u5e7e\u7a2e\u591a\u6587\u4ef6\u6458\u8981\u6280\u8853\u3002MEAD [18]\u63a5\u53d7\u5206\u7fa4\u904e\u5f8c\u7684\u6587\u4ef6\u96c6 1 \uff0c\u4f75\u8003\uf97e\u4ee5\u4e0b\u4e09\u500b\u7279\u5fb5\uff1a 1) \u8a9e\uf906\u8207\u7fa4\u4e2d\u5fc3(Centroid)\u7684\u76f8\u4f3c\ufa01\uff1b2) \u8a9e\uf906\u65bc\u6587\u4ef6\u4e2d\u7684\u4f4d\u7f6e\uff0c\u901a\u5e38\u51fa\u73fe\u65bc\u6587\u4ef6\u9996\uf906\u7684\u8a9e\uf906\uff0c\u53ef \u52a0\u91cd\u5176\u91cd\u8981\u6027\uff1b3) \u8a9e\uf906\u8207\u6240\u5c6c\u6587\u4ef6\u4e4b\u9996\uf906\u7684\u76f8\u4f3c\ufa01\u3002MEAD\u4ee5\u7dda\u6027\u7d44\u5408(Linear Combination)\u7d50\u5408 \u4e0a\u8ff0\u4e09\u7a2e\u7279\u5fb5\uff0c\u7d9c\u5408\u8a55\u4f30\u8a9e\uf906\u91cd\u8981\u7a0b\ufa01\u3002\u4e00\u822c\u800c\u8a00\uff0cMEAD\u4f7f\u7528\u7684\u9996\uf906\u52a0\u91cd\u8a08\u5206\u6cd5\uff0c\u6bd4\u8f03\u9069\u7528\u65bc \u65b0\u805e\u6587\u7ae0 2 \uff1b\u5982\u679c\u6587\u4ef6\u96c6\u662f\u70ba\u5176\u4ed6\uf9b4\u57df\uff0c\uf9b5\u5982\u6280\u8853\uf9d0\u7684\u6587\u4ef6\uff0c\u5247\u9996\uf906\u52a0\u91cd\u8a08\u5206\u6cd5\u8981\u518d\u8abf\u6574\u624d\u5408\u9069\u3002 McKeown et al. [17]\u8a8d\u70ba\u4e3b\u984c\u76f8\u95dc\u7684\u6587\u4ef6\u96c6\u4e2d\uff0c\u5b58\u5728\u6709\u8a31\u591a\uf967\u540c\u7684\u5c0f\u4e3b\u984c(Theme)\u3002\u4ed6\u5011\u7684 \u65b9\u6cd5\uff0c\u5206\u70ba\u4e09\u500b\u90e8\u5206\uff1a1) \u4e3b\u984c\u8fa8\uf9fc(Theme Identification) [8]\u4ee5\u8a9e\uf906\u70ba\u6700\u5c0f\u55ae\u4f4d\uff0c\u900f\u904e\u5206\u7fa4\u6280\u8853\u5c07 \u6587\u4ef6\u4e2d\u7684\u4e3b\u984c\u62bd\u53d6\u51fa\uf92d\uff0c\u540c\u6642\u8fa8\uf9fc\u6587\u4ef6\u9593\u76f8\u4f3c\u53ca\u5dee\uf962\u7684\u90e8\u5206\uff1b2) \u8cc7\u8a0a\u878d\u5408(Information Fusion) [3] \u5c07\u8a0e\uf941\u76f8\u95dc\u4e3b\u984c\u7684\u6bb5\uf918\u878d\u5408\uff0c\u4e26\u53bb\u9664\u91cd\u8907\u7684\u8cc7\u8a0a\uff1b3) \u6458\u8981\u751f\u6210(Text Reformulation) \u5c07\u6240\u6458\uf93f\u51fa \uf92d\u7684\u91cd\u8981\u5b57\u8a5e\u91cd\u65b0\u7d44\u5408\u4ee5\u7522\u751f\uf9ca\u66a2\u7684\u6458\u8981\u3002\u4ed6\u5011\u4e3b\u8981\u8003\u616e\u4ee5\u4e0b\u7279\u5fb5\u4ee5\u6c7a\u5b9a\uf978\u6bb5\uf918\u7684\u76f8\u4f3c\ufa01\uff0c\u9032\u800c \uf9dd\u7528\u5206\u7fa4\u6cd5\u5c07\u627e\u51fa\u4e3b\u984c\uff0c\u5373\u76f8\u4f3c\u6bb5\uf918\u7684\u96c6\u5408\uff1a Word co-occurrence\uff1a\u5047\u5982\uf978\u500b\u6bb5\uf918\u6709\u8a31\u591a\u76f8\u4f3c\u7684\u5b57\uff0c\u5247\u53ef\u8996\u70ba\u76f8\u4f3c\u3002 Matching noun phrases\uff1a\uf9dd\u7528 LinkIt [26]\u5224\u65b7\u662f\u5426\u64c1\u6709\u4e92\u76f8\u95dc\uf997\u7684\u540d\u8a5e\u7247\u8a9e\u7fa4\u7d44\u3002 WordNet synonyms\uff1a\u4f7f\u7528 WordNet [27]\u627e\u51fa\u540c\u7fa9\u8a5e\u7d44\u3002 Common semantic classes for verb\uff1a\u5224\u65b7\u5177\u6709\u540c\u4e00\u8a9e\u610f\u7684\u52d5\u8a5e\u8a5e\u7d44\u3002 \u63a5\u8457\uf9dd\u7528 Information Fusion \u7684\u6280\u8853\uff0c\u5f9e\u4e3b\u984c\u4e2d\u8403\u53d6\u51fa\u5177\u6709\u4ee3\u8868\u6027\u7684\u8a5e\u7d44\u6216\u7247\u8a9e\u3002\u540c\u6642\u4f9d\u7167\u51fa\u73fe \u5728\u6587\u7ae0\u4e2d\u7684\u6b21\u5e8f\uff0c\u5c0d\u7247\u8a9e\u6392\u5e8f\u3002\u6700\u5f8c\uff0c\u85c9\u7531 FUF/SURGE [9]\u81ea\u7136\u8a9e\u8a00\u7522\u751f\u5668\u751f\u6210\u5b8c\u6574\u8a9e\uf906\u3002 MMR (Maximal Marginal Relevance) [4]\u9069\u7528\u65bc\u55ae\u6587\u4ef6\u6458\u8981\uff0c\u5176\u6982\uf9a3\u4e43\u662f\u5c0d\u6240\u6311\u9078\u51fa\u8207 Query \u7684\u6982\uf9a3\uff0c\u53ef\u6709\u6548\ufa09\u4f4e\u6458\u8981\u4e2d\u5177\u6709\u76f8\u540c\u6db5\u7fa9\u7684\u8a9e\uf906(\u5373\uff0c\u6e1b\u5c11\u91cd\u8907\u6027\u8cc7\u8a0a)\u3002MMR-MD \u540c\u6642\u8003\u616e\u5230 \u8a72\u91cd\u8981\u6027\u53ef\u4f5c\u70ba\u6458\u8981\u8a9e\uf906\u6311\u9078\u7684\u4f9d\u64da\u3002 \u5728\u6700\u597d\u7684\u60c5\u6cc1\u4e0b\uff0c\u52a0\u5165\u8a9e\u610f\u7db2\uf937\u53ef\u4ee5\u6bd4\u6c92\u6709\u52a0\u5165\u8a9e\u610f\u7db2\uf937\u6539\u5584\u7d04 7%\u3002\u9019\u500b\uf969\u64da\u986f\u793a\u51fa\u9069\u7576\u5730\u52a0 \u7c21\u55ae\u7684\uf96f\uff0c\u7b2c\u4e00\u7a2e\u65b9\u6cd5\u53ea\u55ae\u7d14\u5224\u65b7\u8a9e\uf906\u4e2d\u6709\u591a\u5c11\u5b57\u51fa\u73fe\u5728\u8a72\u6982\uf9a3\u7fa4\uf9e8\uff0c\u6982\uf9a3\u7fa4\u4e2d\u539f\u672c\u53ea\u5305 defense\u3001abuse\u3001failure\u3001stroke\u3001war\u3001murder \u53ca fire\u3002 \u8a9e\uf906\u65bc\u6587\u4ef6\u4e2d\u51fa\u73fe\u7684\u4f4d\u7f6e \u61c9\u8a72\u8981\u6bd4\u5176\u4ed6\u9060\uf9ea\u4e2d\u5fc3\u9ede\u7684\u6982\uf9a3\u7fa4\u8981\u91cd\u8981\uff0c\u4ea6\u80fd\u52a0\u5f37\u6db5\u84cb\u6027\u8d8a\u5927\u7684\u6982\uf9a3\u7fa4\u91cd\u8981\u6027\u3002 \uf969\u6700\u9ad8\uff0c\u56e0\u6b64\u4e4b\u5f8c\u6703\u5206\u5225\uf9dd\u7528 5\u300120 \u4f5c\u70ba\u5206\u7fa4\u7684\uf969\uf97e\u3002\u5716 5 \u70ba\u8abf\u6574\u8a9e\u610f\u7db2\uf937\u95dc\u4fc2\u6b0a\u91cd\u7684\u7d50\u679c\uff0c \u76f8\u95dc\u7684\u8a9e\uf906\u91cd\u65b0\u6392\u5e8f\uff0c\u4ee5\u7b26\u5408\u5177\u6709\u6700\u5927\u76f8\u95dc\ufa01\u53ca\u6700\u5927\u5dee\uf962\ufa01\u7684\u7279\u6027\u3002MMR-MD [10]\u5ef6\u4f38 MMR \u516c\u5f0f 1: MMR-MD [10] \u5176\u4e2d\uff0cSim 1 MMR-MD \u76ee\u7684\u5728\u65bc\u4f7f\u6458\u8981\u4e2d\u7684\u6bb5\uf918\u5118\u53ef\u80fd\u7684\u76f8\u4f3c\u65bc Query\uff0c\u4f46\u5176\u6240\u9078\u5230\u7684\u6bb5\uf918\u9593\u8981\u5118\u53ef\u80fd \u7684\uf967\u76f8\u4f3c\u3002\u7531\u65bc\u8207 Query \u76f8\u4f3c\ufa01\u9ad8\u7684\u6bb5\uf918\uff0c\u5f7c\u6b64\u4e4b\u9593\u7684\u91cd\u8907\u6027\u53ef\u80fd\u4e5f\u9ad8\uff0c\u800c\u8207\u6bb5\uf918\u76f8\u4f3c\ufa01\u7a0d\u4f4e\u7684 \u6bb5\uf918\uff0c\u5f7c\u6b64\u4e4b\u9593\u7684\u91cd\u8907\u6027\u53ef\u80fd\u4e5f\u4f4e\uff0c\u900f\u904e\u9069\u7576\u7684 \u03bb \u503c\u53ef\u4ee5\u627e\u5230\u517c\u5177\u4e3b\u984c\u4f46\u53c8\uf967\u6703\u6709\u904e\u591a\u91cd\u8907\u6027\u7684 \u6bb5\uf918\u70ba\u6458\u8981\u3002 Mani et al. [15]\u5c07\u6587\u4ef6\u8868\u793a\u6210\u5716\u5f62(Graph)\uff0c\u5176\u4e2d\uff0c\u6bcf\u500b\u7bc0\u9ede\u4ee3\u8868\u4e00\u500b\u95dc\u9375\u8a5e(Term)\uff0c\u7bc0\u9ede\u8207 \u7bc0\u9ede\u9593\u7528\uf967\u540c\u7684\u95dc\u4fc2\uf99a\u63a5\u8d77\uf92d\uff0c\u5305\u542b 1) \u7247\u8a9e\u95dc\u4fc2(PHRASE)\uff1b2) \u5f62\u5bb9\u8a5e\u95dc\u4fc2(ADJ)\uff1b3) \u540c\u7fa9\u95dc \u4fc2(SAME)\uff1b4) \u95dc\uf997\u95dc\u4fc2(COREF)\u3002\u9996\u5148\uff0c\u8ce6\u4e88\u6bcf\u500b\u7bc0\u9ede\u4e00\u6b0a\u91cd(Weight)\uff0c\u6b0a\u91cd\u503c\u521d\u59cb\u70ba\u8a72\u95dc\u9375 \u8a5e\u7684 TF-IDF \u503c\u3002\u63a5\u8457\uff0c\uf9dd\u7528 Spreading Activation \u6f14\u7b97\u6cd5\uff0c\u900f\u904e\u7bc0\u9ede\u9593\u76f8\uf99a\u7684\uf99a\u7d50\u6b0a\u91cd\u8b8a\uf901\u7bc0\u9ede \u7684\u6b0a\u91cd\u503c\uff0c\u4ee5\u627e\u51fa\u8207 Query \u76f8\u95dc\u7684\u7bc0\u9ede\u3002\u63a5\u8457\uff0c\u6bd4\u8f03\uf978\uf978\u6587\u4ef6\u5716\u5f62\u6a21\u578b\u7684\u76f8\u4f3c\ufa01(Commonality) \u53ca\u5dee\uf962\u6027(Difference)\u3002\u4ed6\u5011\u63d0\u51fa FSD (Find Similarities and Differences)\u6f14\u7b97\u6cd5\uff0c\u4ee5\u627e\u51fa\uf978\u5716\u5f62\u4e2d \u76f8\u4f3c\u6216\u5dee\uf962\u7684\u7bc0\u9ede\u3002\u6700\u5f8c\uff0c\u900f\u904e\u5206\u6790 Similarities \u53ca Differences \u96c6\u5408\u4e2d\u7684\u95dc\u9375\u8a5e\uff0c\u8a08\u7b97\u8a9e\uf906\u7684\u91cd \u8981\u6027\uff0c\u4e26\u6311\u9078\u91cd\u8981\u7684\u8a9e\uf906\u70ba\u6458\u8981\u7d50\u679c\u3002 3. \u6982\uf9a3\u5206\u7fa4\u53ca\u62bd\u53d6 \u672c\u7bc0\uf96f\u660e\u5982\u4f55\u4ee5\u7d71\u8a08\u65b9\u6cd5\u53ca\u5206\u7fa4\u6280\u8853\u7531\u65b0\u805e\u6587\u4ef6\u96c6\u4e2d\u63a8\u5c0e\u51fa\u4e8b\u4ef6\u6982\uf9a3\u7fa4 3 \u3002\u9996\u5148\u4ecb\u7d39\u5982\u4f55\u9078 \u53d6\u91cd\u8981\u7684\u6982\uf9a3\u8a5e\uff0c\u4e26\uf9dd\u7528\u6982\uf9a3\u8a5e\u7684\u524d\u5f8c\u6587(Context)\u8207\u8a9e\u610f\u7db2\uf937(Semantic Network)\u4f5c\u70ba\u5176\u63cf\u8ff0\uff1b\u63a5 \u8457\uf96f\u660e\uf9dd\u7528\u5206\u7fa4\u6cd5\u5c07\u6982\uf9a3\u8a5e\u5206\u7fa4\uff0c\u4ee5\u5c0e\u51fa\u65b0\u805e\u4e8b\u4ef6\u4e2d\u7684\u4e3b\u984c\u3002 \u5716 1 \uf96f\u660e\u672c\u6587\u6240\u63d0\u65b9\u6cd5\u4e4b\u67b6\u69cb\u3002\u6b65\u9a5f\u4e00\u70ba\u524d\u7f6e\u8655\uf9e4\uff1b\u6b65\u9a5f\u4e8c\u6311\u9078\u5177\u6709\u4ee3\u8868\u6027\u7684\u540d\u8a5e\u53ca\u540d\u8a5e \u7247\u8a9e\u7576\u4f5c\u5019\u9078\u7684\u6982\uf9a3\u8a5e\uff1b\u6b65\u9a5f\u4e09\u6839\u64da\u5019\u9078\u6982\uf9a3\u8a5e\u4e4b\u524d\u5f8c\u6587\u53ca\u4e8b\u5148\u5efa\uf9f7\u7684\u8a9e\u610f\u7db2\uf937\uff0c\u63cf\u8ff0\u8a72\u5019\u9078\u6982 \uf9a3\u8a5e\uff0c\u4ee5\u5f97\u5230\u4e00\u5411\uf97e\u8868\u793a\u5f0f\uff1b\u6b65\u9a5f\u56db\u91dd\u5c0d\u5019\u9078\u6982\uf9a3\u8a5e\u4f5c\u5206\u7fa4\uff0c\u53ef\u5f97\u5230\u6982\uf9a3\u76f8\u4f3c\u4e4b\u6982\uf9a3\u7fa4\uff1b\u6b65\u9a5f\u4e94 \u4f9d\u64da\u8a9e\uf906\u4e2d\u95dc\u9375\u8a5e\u7684\u8cc7\u8a0a\uff0c\u5c07\u8a9e\uf906\u5c0d\u61c9\u5230\u6982\uf9a3\u7fa4\u4e2d\uff0c\u5f97\u5230\u8a9e\uf906\u8207\u6982\uf9a3\u7fa4\u7684\u95dc\uf99a\uff1b\u6b65\u9a5f\uf9d1\u6839\u64da\u8a9e\uf906 \u8207\u6982\uf9a3\u7fa4\u7684\u95dc\uf997\uff0c\u540c\u6642\u8003\uf97e\u6587\u7ae0\u7d50\u69cb\u7684\u95dc\u4fc2\uff0c\u8a08\u7b97 3 \u500b\u8207\u8a9e\uf906\u76f8\u95dc\u53ca 2 \u500b\u8207\u6982\uf9a3\u7fa4\u76f8\u95dc\u7684\u7279\u5fb5\u503c\uff1b \u6b65\u9a5f\u4e03\u5247\u4f9d\u64da\u6b65\u9a5f\uf9d1\u8a08\u7b97\u4e4b\u7279\u5fb5\u503c\uff0c\u4ee5\u7dda\u6027\u7d44\u5408\u7684\u65b9\u5f0f\uff0c\u8a08\u7b97\u4f4d\u65bc\u540c\u4e00\u6982\uf9a3\u7fa4\u4e2d\u8a9e\uf906\u7684\u91cd\u8981\u6027\uff0c \u5716 1: \u7cfb\u7d71\u67b6\u69cb\u5716 3.1. \u7d50\u5408\u524d\u5f8c\u6587\u8207\u8a9e\u610f\u7db2\uf937\u7684\u6982\uf9a3\u63cf\u8ff0\u6cd5 \u9996\u5148\uff0c\u9032\ufa08\u524d\u7f6e\u8655\uf9e4(Preprocessing)\uff0c\u5176\u4f5c\u7528\u70ba\u907f\u514d\u96dc\u8a0a\u5e72\u64fe\uff0c\ufa09\u4f4e\u7d71\u8a08\uf969\u64da\u7684\uf96b\u8003\u6027\u3002\u6b64 \u6b65\u9a5f\u5305\u542b\u65b7\u8a5e\ufa00\u5b57(Tokenization)\u3001\u8a5e\u6027\u6a19\u8a18(Part-Of-Speech)\u3001\u8a5e\u5e79\u9084\u539f(Stemming)\u3001\u5c0f\u5beb\u5316 (Lowercasing)\u3001\u522a\u9664\u505c\u7528\u5b57(Stopword Removing)\u53ca\u7247\u8a9e\u5316(Chunking)\u7b49\u3002\u672c\u6587\u4e2d\uff0c\u65b7\u8a5e\ufa00\u5b57\u53ca\u8a5e \u6027\u6a19\u8a18\u63a1\u7528 NLP Processor [20]\uff1b\u8a5e\u5e79\u9084\u539f\u63a1\u7528 Porter \u6f14\u7b97\u6cd5[24]\uff1b\u7247\u8a9e\u5316\u5247\uf9dd\u7528\u7d71\u8a08\u65b9\u6cd5\u8a08\u7b97\u8a5e \u6027\u7d44\u5408\u6a5f\uf961\uff0c\u4ee5\u8fa8\u5225\u662f\u5426\u70ba\u53ef\u7d44\u5408\u7247\u8a9e\uff1b\u505c\u7528\u5b57\u7684\u90e8\u4efd\u91dd\u5c0d DUC 2003 \u6240\u63d0\u4f9b\u7684\u6587\u4ef6\u96c6\u8a2d\u8a08\uff0c\u5171 \u6709 309 \u500b\u5b57\uff0c\u5176\u4e2d\u7d55\u5927\u591a\uf969\u70ba\u4ecb\u4fc2\u8a5e\u3001\u6307\u4ee3\u8a5e\u3001\uf99a\u8a5e\u53ca\u52a9\u8a5e\u3002 \u524d\u7f6e\u8655\uf9e4\u5f8c\u50c5\u4fdd\uf9cd\u540d\u8a5e(Noun)\u53ca\u540d\u8a5e\u7247\u8a9e(Noun Phrase)\u4f5c\u70ba\u53ef\u80fd\u7684\u5019\u9078\u6982\uf9a3(Concept Candidate)\uff0c\u539f\u56e0\u4e43\u662f\u540d\u8a5e\u6bd4\u5176\u4ed6\u8a5e\u6027\u542b\u6709\uf901\u591a\u8a9e\u610f[1] [13]\u3002\u672c\uf941\u6587\u540c\u6642\u8a08\u7b97\u6bcf\u500b\u5019\u9078\u8a5e\u7684 tf-idf \u503c[28]\uff0c\u9032\u4e00\u6b65\u904e\uf984\uf967\u5177\u4ee3\u8868\u6027\u7684\u5b57\u8a5e\u3002\u6700\u5f8c\uff0c\u518d\u7531\u6982\uf9a3\u5019\u9078\u8a5e\u4e2d\u6311\u9078\u4e00\u822c\u540d\u8a5e\u3001\u8907\uf969\u540d\u8a5e\u3001\u5c08 \u6709\u540d\u8a5e\u3001\u8907\uf969\u5c08\u6709\u540d\u8a5e\u7b49\u5b57\u8a5e\uff0c\u4f5c\u70ba\u6700\u5f8c\u6240\u4fdd\uf9cd\u7684\u6982\uf9a3\u5019\u9078\u8a5e\u96c6\u5408\u3002 \u63a5\u8457\uf96f\u660e\u5982\u4f55\u5c0e\u51fa\u6982\uf9a3\u5019\u9078\u8a5e\u7684\u8868\u793a\u6cd5\u3002[2]\u63d0\u5230\u7d55\u5927\u591a\uf969\u63cf\u8ff0\u540c\u4e00\u4e8b\u4ef6\u6240\u4f34\u96a8\u51fa\u73fe\u7684\u5b57 \u8a5e\uff0c\u5176\u8a9e\u610f\u7686\u5f88\u76f8\u4f3c\u3002[5]\u4ea6\u63d0\u5230\u9664\u8003\u616e\u55ae\u4e00\u5b57\u8a5e\u7684\u91cd\u8981\u6027\u5916\uff0c\uf901\uf967\u80fd\u5ffd\uf976\u51fa\u73fe\u5728\u91cd\u8981\u8a5e\u5f59\u524d\u5f8c \u6587\u7684\u5f71\u97ff\uf98a\uff1b\uf9b5\u5982\uff0ccondemn(\u8b74\u8cac)\u53ca intensively(\u5f37\uf99f)\u7d93\u5e38\u4e00\u8d77\u51fa\u73fe\uff0c\u6b64\uf978\u95dc\u9375\u8a5e\u53ef\u7528\uf92d\u63cf\u8ff0\u5f7c \u6b64\u3002\u57fa\u65bc\u9019\u500b\u60f3\u6cd5\uff0c\u672c\u7814\u7a76\uf9dd\u7528\u5019\u9078\u6982\uf9a3\u8a5e\u7684\u524d\u5f8c\u6587\u63cf\u8ff0\u8a72\u5b57\u8a5e\u3002\u4f5c\u6cd5\u4e0a\u4ee5\u51fa\u73fe\u5728\u5019\u9078\u6982\uf9a3\u8a5e\u524d \u8868 1: \u4ee5\u524d\u5f8c\u6587\u63cf\u8ff0\u5f8c\u9078\u6982\uf9a3\u7bc4\uf9b5 4 Concept Indexing Description Length the U.S. Embassy several locations, Bonn, receiving, word, a terrorist threat, the U.S. Embassy, no evidence, a planned attack, found, officials, Wednesday 11 \u70ba\uf9ba\u52a0\u5f37\u63cf\u8ff0\u5b57\u8a5e\u7684\u8a9e\u610f\uff0c\u672c\u7814\u7a76\u5728\u4ee5\u524d\u5f8c\u6587\u63cf\u8ff0\u6982\uf9a3\u6642\u4ea6\u52a0\u5165\u8a9e\u610f\u7db2\uf937\u3002\u6211\u5011\u4f7f\u7528 Informap [12]\u5efa\uf9f7\u65b0\u805e\u4e8b\u4ef6\u7684\u8a9e\u610f\u7db2\uf937\uff0cInfomap \u4ee5\u5171\u73fe (Co-occurrence) \u7684\u539f\u5247\uf92d\u5224\u65b7\u5b57\u8207\u5b57\u4e4b \u9593\u8a9e\u610f\u7684\u76f8\u95dc\u6027\u3002\u9996\u5148\uff0c\u4f9d\u7167\u5b57\u51fa\u73fe\u7684\u983b\uf961\u9078\u8a02\u51fa\u8a9e\u610f\u57fa\u672c\u5b57(Content-Bearing Words)\uff0c\u518d\u8a02\u51fa \u4e00\u500b\u53ef\u8abf\u5f0f\u7bc4\u570d(Window)\uff0c\u5728\u9019\u500b\u7bc4\u570d\u5167\u7684\u6bcf\u4e00\u500b\u5b57\u4f34\u96a8\u8a9e\u610f\u57fa\u672c\u5b57\u4e00\u8d77\u51fa\u73fe\u7684\u983b\uf961\uff0c\u5c31\u628a\u9019 \ufa09\u4f4e\u5b57\u5411\uf97e\u7684\u7dad\ufa01\uff0c \u6700\u5f8c\u8a08\u7b97\uf978\uf978\u8a5e\u9593\u7684\u76f8\u4f3c\ufa01\uff0c\u4e26\u5efa\uf9f7\u8a9e\u610f\u7db2\uf937\u3002\u5716 2 \u70ba\u4e00\u4ee5 attack \u70ba\u4e2d\u5fc3 \u7684\u8a9e\u610f\u7db2\uf937\u7bc4\uf9b5\uff0c\u900f\u904e\u8a9e\u610f\u7db2\uf937\uff0c\u53ef\u627e\u51fa\u8207 attack \u8a9e\u610f\u76f8\u95dc\u7684\u5b57\u8a5e\u3002\uf9b5\u5982\uff0c\u76f4\u63a5\u76f8\u95dc\u7684\u5b57\u8a5e\u6709 \u54ea\u500b\u6982\uf9a3\u7fa4\u4e2d\u7684\u5b57\uf969\u6700\u591a\uff0c\u5247\u6b78\uf9d0\u5230\u8a72\u6982\uf9a3\u7fa4\u3002\u7b2c\u4e8c\uff0c\u5224\u65b7\u8a9e\uf906\u4e2d\u7684\u6982\uf9a3\u51fa\u73fe\u5728\u54ea\u500b\u6982\uf9a3\u7fa4\u4e2d\u7684 \u5b57\uf969\u6700\u591a\uff0c\u5247\u6b78\uf9d0\u5230\u8a72\u6982\uf9a3\u7fa4\u3002\u516c\u5f0f 2 \u70ba\u8a9e\uf906\u5c0d\u61c9\u5230\u6982\uf9a3\u7fa4\u7684\u5224\u65b7\u4f9d\u64da\u3002 length sentence IDF TF S m n i i tfidf _ \u239f \u23a0 \u239e \u239c \u5fc3\u9ede\u7684\u7d66\u4e88\u8d8a\u9ad8\u5206\u3002\u5728\u4e2d\u5fc3\u9ede\u9644\u8fd1\u7684\u6982\uf9a3\u7fa4\uff0c\u8d8a\u6709\u53ef\u80fd\u6db5\u84cb\u5176\u4ed6\u6982\uf9a3\u7fa4\u7684\u610f\u601d\uff0c\u5728\u9806\u5e8f\u4e0a \u8b8a\uf969\u7684\u5be6\u4f5c\u8a55\u4f30\uff0c\u7531\u5716 4 \u5f97\u77e5\u5728 5 \u7fa4\u6642 ROUGE-1 \u7684\u5206\uf969\u6700\u9ad8\uff0c\u5728 20 \u7fa4\u7684\u6642\u5019 ROUGE-L \u7684\u5206 \u239d \u00d7 = = \u2211 \u5206\u7fa4\u7684\u7d50\u679c\uff0c\u4f9d\u7167\u5411\uf97e\u7684\u5206\u4f48\u60c5\u5f62\u53ef\u4ee5\u627e\u51fa\u5168\u90e8\u5411\uf97e\u7684\u4e2d\u5fc3\u9ede\u3002\u6bcf\u4e00\u500b\u6982\uf9a3\u7fa4\u4e2d\u8d8a\u9760\u8fd1\u4e2d \u5b57\u5f59\uff0c\u9019\u5340\u9593\u7684\u5b57\u4e5f\u6700\u70ba\u76f8\u4f3c\uff0c\u5be6\u9a57\u7d50\u679c\u4e5f\u6bd4\u5176\u4ed6\u8d85\u904e\u5340\u9593\u7684\u9577\ufa01\u70ba\u9ad8\u3002\u5716 4 \u70ba\u8abf\u6574\u5206\u7fa4\uf969\uf97e \u239b \u4e9b\u983b\uf961\u5b9a\u5728\u5171\u73fe\u77e9\u9663\u3002Infomap \uf9dd\u7528\u5171\u73fe\u77e9\u9663\uff0c\u4e26\u4f7f\u7528\u5947\uf962\u503c\u5206\u89e3(Singular Value Decomposition) \u8868 2: \u52a0\u5165\u8a9e\u610f\u7db2\uf937\u65b9\u6cd5 1 \u7684\u7bc4\uf9b5 Concept Indexing Description Length the U.S. Embassy several locations, receiving, a terrorist threat, the U.S. Embassy, no evidence, a planned attack 6 \u7b2c\u4e8c\u7a2e\u65b9\u6cd5\u5247\u662f\u5e0c\u671b\u80fd\u5920\u7a81\u986f\u8207\u6982\uf9a3\u8a9e\u610f\u6709\u95dc\u7684\u5b57\u8a5e\uff0c\u4e14\uf967\u81f3\u65bc\u5f71\u97ff\u5230\u539f\u672c\u63cf\u8ff0\u5b57\u8a5e\u7684\u7d44 \u6210\uff0c\u6211\u5011\u4fdd\uf9cd\uf9ba\u539f\u59cb\u7528\u4ee5\u63cf\u8ff0\u6982\uf9a3\u7684\u6240\u6709\u524d\u5f8c\u6587\u5b57\u5f59\uff0c\u4f46\u52a0\u91cd\u5728\u8a9e\u610f\u7db2\uf937\u4e0a\u8207\u6982\uf9a3\u76f8\u95dc\u7684\u5b57\u5f59\u3002 \u4ee5\u7f8e\u570b\u5927\u4f7f\u9928(the U.S. Embassy)\u6b64\u4e00\u6982\uf9a3\u70ba\uf9b5\uff0c\u7531\u65bc several locations, receiving, a terrorist threat, \u7531\u4e0a\u8ff0\uf978\uf9b5\u53ef\u77e5\uff0c\u65b9\u6cd5\u4e00\u8207\u65b9\u6cd5\u4e8c\u7684\u5dee\u5225\u70ba\uff0c\u65b9\u6cd5\u4e00\u6839\u64da\u8a9e\u610f\u7db2\uf937\u522a\u9664\uf967\u91cd\u8981\u7684\uf96a\u5f15\u8a5e\uff0c\u800c \u65b9\u6cd5\u4e8c\u5247\u662f\u4fdd\uf9cd\u6240\u6709\u7684\uf96a\u5f15\u8a5e\uff0c\u4f46\u52a0\u91cd\u5728\u8a9e\u610f\u7db2\uf937\u4e2d\u8207\u6982\uf9a3\u76f8\u95dc\u8a5e\u4e4b\u6b0a\u91cd\u503c\u3002\u53e6\u5916\uff0c\u65b9\u6cd5\u4e00\u7684\u63cf \u8ff0\u96d6\u7136\u6bd4\u8f03\u80fd\u5920\u8cbc\u8fd1\u6982\uf9a3\u7684\u8a9e\u610f\uff0c\u4f46\u5176\u63cf\u8ff0\u7684\u5b57\u5f59\u5206\u4f48\u6bd4\u8f03\u6563\uff0c\u4e14\u53bb\u9664\uf9ba\uf967\u5728\u8a9e\u610f\u7db2\uf937\u5167\u7684\u5b57 \u5f59\uff0c\u4f7f\u5f97\u63cf\u8ff0\u7684\u5b57\u5f59\uf969\u76ee\u5c11\u65bc\u539f\u672c\u7684\u63cf\u8ff0\u5b57\u5f59\uff0c\u56e0\u6b64\uff0c\u65b9\u6cd5\u4e00\u96d6\u7136\u63cf\u8ff0\u7cbe\u6e96\u4f46\u662f\u6703\u6709\u63cf\u8ff0\u5b57\u5f59\uf967 \u8db3\u7684\u60c5\u5f62\uff0c\uf99a\u5e36\u6703\u5f71\u97ff\u5230\u4e4b\u5f8c\u5728\u5206\u7fa4\u4ee5\u53ca\u5f8c\uf92d\u8a08\u7b97\u7279\u5fb5\u7684\u6b0a\u91cd\u3002 \u8868 3: \u52a0\u5165\u8a9e\u610f\u7db2\uf937\u65b9\u6cd5 2 \u7684\u7bc4\uf9b5 Concept Indexing Description the U.S. Embassy several locations (5.1705+X), Bonn (5.1705), receiving (2.9733+X), word (5.1705), a terrorist threat (5.1705+X), the U.S. Embassy (2.9733+X), no evidence (5.1705+X), a planned attack (5.1705+X), found (3.5611), officials (4.4773), Wednesday (2.9733) \u672c\u6587\u4e2d\u5206\u7fa4\u7684\u5c0d\u8c61\uff0c\u662f\u7d93\u904e 3.1 \u8655\uf9e4\u5f8c\u4e4b\u6982\uf9a3\u5411\uf97e\uff0c\u5206\u7fa4\u65b9\u6cd5\u5247\u63a1\u7528 K-Means [11]\u3002\u8003\uf97e\u65b0 \u805e\u4e8b\u4ef6\u53ef\u518d\u7d30\u5206\u70ba\u5730\u9ede\u3001\u5c0d\u8c61\u3001\u5f71\u97ff\u7d50\u679c\u7b49\u7279\u6027\uff0c\u5206\u7fa4\u7684\u7d50\u679c\u53ef\u8996\u70ba\u6587\u4ef6\u4e2d\u6240\u63d0\u53ca\u7684\u4e3b\u984c\u6982\uf9a3\u3002 \u6982\uf9a3\u5206\u7fa4\u4e4b\u5f8c\uff0c\uf965\u8981\u5c07\u8a9e\uf906\u8207\u6982\uf9a3\u7fa4\u4f5c\uf99a\u7d50\uff0c\u4ee5\u671f\u627e\u5230\u80fd\u5920\u4ee3\u8868\u6bcf\u500b\u8a9e\uf906\u7684\u6982\uf9a3\u7fa4(\u4ea6\u5373\uff0c \u8207[5]\u63d0\u5230\u7684\u8cc7\uf9be\u543b\u5408\u3002\u4ea6\u5373\uff0c\u4f9d\u7167\u4eba\uf9d0\u66f8\u5beb\u4ee5\u53ca\u95b1\uf95a\u7fd2\u6163\u5728\u770b\u5230\u67d0\u500b\u5b57\u6642\uff0c\u6703\u8a18\u61b6\u5230\u524d 7\u00b12 \u500b \u8a72\u8a9e\uf906\u6240\u8981\u8868\u9054\u7684\u8a9e\u610f\u53ca\u76f8\u95dc\u4e3b\u984c)\u3002\u672c\u6587\u63d0\u51fa\uf978\u7a2e\u5c0d\u61c9\u65b9\u6cd5\u3002\u7b2c\u4e00\uff0c\u5224\u65b7\u8a9e\uf906\u4e2d\u7684\u5b57\u8a5e\u51fa\u73fe\u65bc (1)SIM s,i = Words Match = sim ( Match_Word, Cluster j ) / L_of_S (2)SIM s,i = Concepts Match = sim ( Match_Vector, Cluster j ) / L_of_S Match_Vector\uff1aconcept vector included in this sentence sim ( Match_Word, Cluster j )\uff1anumber of word appear in cluster j sim ( Match_Vector, Cluster j )\uff1adistance between vector and centroid of cluster j L_of_S\uff1alength of the sentence \u516c\u5f0f 2: \u8a9e\uf906\u5c0d\u61c9\u5230\u6982\uf9a3\u7fa4\u7684\u65b9\u5f0f \u6bd4\u8f03\u4e0a\u8ff0\uf978\u500b\u65b9\u6cd5\uff0c\u65b9\u6cd5\u4e00\u7684\u5c0d\u61c9\u7531\u65bc\u628a\u63cf\u8ff0\u6982\uf9a3\u7684\u5b57\u5f59\u4e5f\u52a0\u5165\u5c0d\u61c9\u7684\u689d\u4ef6\uff0c\u56e0\u6b64\u5e7e\u4e4e\u6587 \u4ef6\u96c6\u5167\u7684\u6bcf\u4e00\u500b\u8a9e\uf906\u90fd\u53ef\u4ee5\u627e\u5230\u5c0d\u61c9\u5230\u7684\u6982\uf9a3\u7fa4\uff0c\u9020\u6210\uf9ba\u6bcf\u4e00\u500b\u6982\uf9a3\u7fa4\u5167\u7684\u8a9e\uf906\uf969\uf97e\u591a\uff0c\u4f46\u662f\u8a9e \uf906\u7684\u8a9e\u610f\u53ef\u80fd\uf967\u662f\u8207\u6982\uf9a3\u7fa4\u7684\u6982\uf9a3\u9ad8\ufa01\u76f8\u4f3c\uff0c\u9020\u6210\u6b64\u73fe\u8c61\u7684\u539f\u56e0\u53ef\u80fd\u70ba\u53ea\u5c0d\u61c9\u5230\u63cf\u8ff0\u6982\uf9a3\u7684\u5b57 \u5f59\uff0c\u4e26\uf967\u662f\u5c0d\u61c9\u5230\u6982\uf9a3\u672c\u8eab\u3002\u65b9\u6cd5 2 \u7684\u5c0d\u61c9\u5247\u53ef\u4ee5\u6709\u6548\u7684\u904e\uf984\u6389\u8a9e\u610f\uf967\u7b26\u5408\u6982\uf9a3\u7fa4\u7684\u8a9e\uf906\uff0c\u96d6\u7136 \u5c0d\u61c9\u5f8c\u6bcf\u500b\u6982\uf9a3\u7fa4\u5305\u6db5\u84cb\u7684\u8a9e\uf906\uf969\u76ee\u8f03\u5c11\uff0c\u96d6\u7136\u5269\u4e0b\u7684\u8a9e\uf906\uf969\uf97e\u8f03\u5c11\uff0c\u4f46\u662f\u518d\u7d93\u7531\u5f8c\u9762\u7684\u7279\u5fb5\u9078 \u53d6\u6642\uff0c\u4ea6\u53ef\u6709\u6548\u63d0\u5347\u9078\u53d6\u9069\u5408\u6458\u8981\u8a9e\uf906\u7684\u6548\uf961\u3002 4. \u8a9e\uf906\u8a9e\u610f\u6b0a\u91cd\u6458\u8981 \u900f\u904e\u6982\uf9a3\u7fa4\u53ca\u5176\u4e2d\u6982\uf9a3\u5b57\u8a5e\u8207\u8a9e\uf906\u95dc\u9375\u8a5e\u7684\u76f8\u4f3c\u95dc\u4fc2\uff0c\u53ef\u5c07\u6bcf\u500b\u8a9e\uf906\u5c0d\u61c9\u81f3\u8a9e\u610f\u76f8\u8fd1\u7684\u6982 \uf9a3\u7fa4\u3002\u7136\u800c\uff0c\u4f4d\u65bc\u540c\u4e00\u6982\uf9a3\u7fa4\u4e2d\u7684\u8a9e\uf906\u5f7c\u6b64\u8a9e\u610f\u8fd1\u4f3c\uff0c\u4ecd\u9700\u8981\u900f\u904e\u5176\u4ed6\u689d\u4ef6\u4ee5\u5224\u65b7\u54ea\u500b\u8a9e\uf906\u6700\u80fd \u4ee3\u8868\u8a72\u6982\uf9a3\u7fa4\u3002\u7531\u8a9e\uf906\u7279\u5fb5\u6311\u9078\u91cd\u8981\u8a9e\uf906\u7684\u65b9\u6cd5\u5728\u5f88\u591a\u7814\u7a76\u4e2d\u88ab\u63d0\u51fa\uf92d\uff0c\u85c9\u7531\u62bd\u53d6\uf967\u540c\u7684\u7279\u5fb5\uff0c \u53ef\u4ee5\u6574\u5408\u9019\u4e9b\u7279\u5fb5\u4ee5\u5224\u65b7\u8a9e\uf906\u7684\u91cd\u8981\u7a0b\ufa01[16]\u3002\u672c\u6587\u8003\uf97e 3 \u500b\u8207\u8a9e\uf906\u76f8\u95dc\u53ca 2 \u500b\u8207\u6982\uf9a3\u7fa4\u76f8\u95dc\u7684 4.1. \u8a9e\uf906\u76f8\u95dc\u7279\u5fb5 TF*IDF \u8003\u616e\u8a9e\uf906\u4e2d\u6240\u6709\u5b57\u8a5e\u7684 TF*IDF \u7e3d\u548c\uff0c\u4e26\u9664\u4ee5\u8a9e\uf906\u9577\ufa01\u4ee5\u6b63\u898f\u5316(Normalization)\u3002 \u65b9\u6cd5\u4e00\uff1a\u63a1\u7528\u6240\u5c0d\u61c9\u5230\u7684\u5b57\u5f59\uf969\u76ee\u8a08\u7b97\u76f8\u4f3c\ufa01\uff0c\u4e26\u9664\u4ee5\u8a9e\uf906\u9577\ufa01\u4f5c\u6b63\u898f\u5316\uff0c\u5982\u516c\u5f0f 3\u3002 \u7136\u800c\uff0c\u7531\u5be6\u9a57\u4e2d\u767c\u73fe\u4ee5\u9019\u7a2e\u65b9\u5f0f\uf92d\u8a08\u7b97\u76f8\u4f3c\ufa01\uff0c\u6703\u767c\u751f\u6709\u5f88\u591a\u8a9e\uf906\u6240\u5c0d\u61c9\u5230\u7684\u5b57\u5f59\uf969\uf97e\u662f \u4e00\u6a23\u7684\u60c5\u5f62\uff0c\u5c0e\u81f4\u9019\u500b\u65b9\u6cd5\u6240\u8a08\u7b97\u51fa\u7684\u6b0a\u91cd\uf967\u5177\u6709\u8fa8\u5225\u6027\u3002 length sentence words match S i sim _ / _ = match_words: \u8a08\u7b97\u8207\u6982\uf9a3\u7fa4 i \u5167\u6709\u591a\u5c11\u5b57\u5f59\u662f\u4e00\u6a23\u7684 i: \u8a9e\uf906\u6240\u5c0d\u61c9\u5230\u7684\u6982\uf9a3\u7fa4 i sentence_length: \u8a9e\uf906\u7684\u9577\ufa01 \u516c\u5f0f 3: \u76f8\u4f3c\ufa01\u7279\u5fb5\u8a08\u7b97\u65b9\u6cd5 1 \u65b9\u6cd5\u4e8c\uff1a\u76f8\u4f3c\ufa01\u7684\u8a08\u7b97\u53d6\u6c7a\u65bc\u5411\uf97e\u5c0d\u61c9\u7684\u6982\uf9a3\u7fa4\u8207\u5176\u4e2d\u5fc3\u9ede\u7684\u8ddd\uf9ea\uff0c\u5982\u516c\u5f0f 4\u3002\u6b64\u65b9 \u6cd5\u53ef\u6bd4\u8f03\u54ea\u4e9b\u8a9e\uf906\u6bd4\u8f03\u63a5\u8fd1\u8a72\u6982\uf9a3\u7fa4\u7684\u4e2d\u5fc3\u9ede\u3002\u5728\u591a\u7dad\ufa01\u7684\u5411\uf97e\u4e2d\uff0c\u4f7f\u7528\u6b50\u57fa\uf9e4\u5f97\u8ddd\uf9ea (Euclidean Distance)\u53ef\uf901\u7cbe\u78ba\u5730\u627e\u51fa\u54ea\u4e9b\u5411\uf97e\u63a5\u8fd1\u4e2d\u5fc3\u9ede\uff0c\u6bcf\u500b\u8a9e\uf906\u5c07\u53ef\u4ee5\uf901\u6e05\u695a\u5730\u5206\u51fa\u4ee3 \u8868\u6982\uf9a3\u7fa4\u7684\u91cd\u8981\u6027\u3002 ( ) s s i j i j s L cluster concept ce dis SIM \u00d7 = \u2211 \u2282 ) , ( tan 1 , concept: \u8a9e\uf906\u6709\u5c0d\u61c9\u5230\u6982\uf9a3\u7fa4\u7684\u6982\uf9a3 distance(concept, cluster): \u53d6\u5411\uf97e\u5230\u6982\uf9a3\u7fa4\u4e2d\u5fc3\u7684\u8ddd\uf9ea L s : \u8a9e\uf906\u9577\ufa01 \u516c\u5f0f 4: \u76f8\u4f3c\ufa01\u7279\u5fb5\u8a08\u7b97\u65b9\u6cd5 2 4.2. \u6982\uf9a3\u7fa4\u76f8\u95dc\u7279\u5fb5 \u6982\uf9a3\u7fa4\u5167\u542b\u7684\u6982\uf9a3\u591a\u5be1 \u5305\u542b\u8d8a\u591a\u7684\u6982\uf9a3\uf969\uf97e\uff0c\u8868\u793a\u539f\u6587\u4ef6\u96c6\u63d0\u5230\u7684\u8a31\u591a\u6982\uf9a3\u90fd\u5728\u540c\u4e00\u500b\u7fa4\u3002\u7576\u5305\u542b\u8d8a\u591a\u6982\uf9a3\u7684\u7fa4\uff0c \u5176\u6b0a\u91cd\u61c9\u8a72\u8d8a\u9ad8[5]\u3002 \u6982\uf9a3\u7fa4\u8207\u4e2d\u5fc3\u9ede\u7684\u8ddd\uf9ea \u516c\u5f0f 5: \u8a08\u7b97\u8a9e\uf906\u6b0a\u91cd\u7e3d\u548c\u516c\u5f0f 5. \u5be6\u9a57\u7d50\u679c\u5206\u6790\u8207\u8a55\u4f30 \u81ea\u52d5\u6458\u8981\u7684\u6210\u6548\u8a55\u4f30\uff0c\u53ef\u5206\u70ba\u76f4\u63a5(Intrinsic)\u8207\u9593\u63a5(Extrinsic)\u8a55\u4f30\uf978\u7a2e\u65b9\u5f0f[16]\u3002\u76f4\u63a5\u8a55\u4f30 \u9700\u5148\u5b9a\u7fa9\u51fa\u4e00\u7d44\uf9e4\u60f3\u7684\u6458\u8981\u6e96\u5247\u6216\u7b54\u6848\uff0c\u7136\u5f8c\u8ddf\u7cfb\u7d71\u53d6\u51fa\u7684\u6458\u8981\u505a\u6bd4\u8f03\u3002\u9593\u63a5\u7684\u65b9\u5f0f\u5247\u7121\u9808\u5177\u5099 \uf9e4\u60f3\u7684\u6458\u8981\u7b54\u6848\uff0c\u800c\u662f\u8a55\u4f30\u81ea\u52d5\u6458\u8981\u7684\u7d50\u679c\u5728\u5176\u4ed6\u76f8\u95dc\u61c9\u7528\u7684\u6210\u6548\u3002\u672c\u6458\u8981\u7cfb\u7d71\u4f7f\u7528\u7684\u6e2c\u8a66\u6587\u4ef6 \u96c6\u70ba DUC 2003 (Document Understanding Conferences 2003) [6]\uff0c\u6587\u4ef6\u5167\u5bb9\u662f\u82f1\u6587\u7684\u65b0\u805e\u6587\u4ef6\uff0c\u5206 \u6210 30 \u500b\u65b0\u805e\u4e8b\u4ef6\uff0c\u6bcf\u500b\u4e8b\u4ef6\u4e2d\u7d04\u6709 10 \u7bc7\u76f8\u540c\u4e3b\u984c\u7684\u65b0\u805e\uff0cDUC 2003 \u4e26\u8acb\uf967\u540c\u7684\u5c08\u5bb6\u5c0d\u540c\u4e00\uf9d0 \u5225\u4f5c\u4e09\u7bc7\u6458\u8981\u3002\u8a55\u4f30\u65b9\u6cd5\u662f\u5c07\u7cfb\u7d71\u81ea\u52d5\u7522\u751f\u7684\u6458\u8981\u8207 DUC 2003 \u7684\u5c08\u5bb6\u6240\u4f5c\u51fa\u7684\u6458\u8981\u6bd4\u8f03\uff0c\u6bcf\u500b \u4e8b\u4ef6\u7684\u6458\u8981\u4ee5 100 \u5b57\u70ba\u4e0a\u9650\u3002\u6548\u80fd\u8a55\u4f30\u63a1\u7528 ROUGE (Recall-Oriented Understudy for Gisting Evaluation) [22]\uff0c\u4e3b\u8981\u6bd4\u8f03\u7684\u9805\u76ee\u70ba ROUGE-N\u3001ROUGE-L\uff0c\u5206\u5225\u4ee3\u8868\u300c\u81ea\u52d5\u6458\u8981\u6709\u591a\u5c11 N \u5b57\u8a5e \u8207\u4eba\u5de5\u6458\u8981\u4e00\u6a23\u300d\u53ca\u300c\u81ea\u52d5\u6458\u8981\u8207\u4eba\u5de5\u6458\u8981\u6709\u591a\u5c11\u5b57\u5f59\u662f\u51fa\u73fe\u5728\u540c\u4e00\u8a9e\uf906\u300d\u3002 \u672c\u6587\u6240\u63d0\u51fa\u7684\u65b9\u6cd5\u6709\u8af8\u591a\u8b8a\uf969\u9700\u8981\u6700\u4f73\u5316\u4ee5\u8abf\u6574\u7cfb\u7d71\u6548\u80fd\uff0c\u8868 4 \uf99c\u51fa\u6240\u6709\u53ef\u80fd\u7684\u8b8a\uf969\u3002 \u8868 4: \u5be6\u9a57\u8b8a\uf969\uf96f\u660e \u6b65\u9a5f \u8b8a\uf969 \uf96f\u660e \u524d\u5f8c\u6587\u9577\ufa01 \u63cf\u8ff0\u5b57\u5f59\u7684\u9577\ufa01(\u5373 N \u8207 M) \u63cf\u8ff0\u6982\uf9a3 \u52a0\u5165\u8a9e\u610f\u7db2\uf937\u65b9\u6cd5 \u52a0\u5165\u8a9e\u610f\u7db2\uf937\u5f8c\u63cf\u8ff0\u6982\uf9a3\u5b57\u5f59\u7684\u65b9\u6cd5 \u5206\u7fa4\uf969\uf97e K-means \u5206\u7fa4\u6cd5\u9700\u8981\u5148\u8a2d\u5b9a K \u503c \u5206\u7fa4 \u8a9e\uf906\u5c0d\u61c9\u6982\uf9a3\u7fa4\u65b9\u6cd5 \u5982\u4f55\u5224\u65b7\u8a9e\uf906\u6240\u5c6c\u7684\u6982\uf9a3\u7fa4 \u8a9e\uf906\u91cd\u8981\u6027 \u6b0a\u91cd\u6bd4\uf9b5\u8abf\u6574 \u4e94\u500b\u7279\u5fb5\u4ee5\u4f55\u7a2e\u6bd4\uf9b5\u8a08\u7b97\u624d\u80fd\u6311\u51fa\u6700\u9069\u7576\u7684\u8a9e\uf906 \u6bd4\uf9b5\u3002\u8abf\u6574\u7684\u65b9\u6cd5\u662f\u5148\u8b8a\u63db\u4e00\u500b\u8b8a\uf969\uff0c\u540c\u6642\u56fa\u5b9a\u5176\u4ed6\u56db\u500b\u3002\u6700\u5f8c\u6211\u5011\u6240\u8abf\u6574\u51fa\u8a08\u7b97\u8a9e\uf906\u91cd\u8981\u6027\u4e4b \u7279\u5fb5\u6b0a\u91cd\u6bd4\uf9b5\u03b1:\u03b2:\u03b3:\u03b8:\u03bb\u70ba 1:5:5:1:8\u3002 \u5716 3 \u8abf\u6574\u7684\u8b8a\uf969\u662f\u6982\uf9a3\u5411\uf97e\u7684\u9577\ufa01\uff0c\u4e5f\u5c31\u662f\u7528\uf92d\u63cf\u8ff0\u6982\uf9a3\u6240\u7528\u7684\u524d\u5f8c\u6587\u9577\ufa01\u3002\u8a55\u4f30\u7684\u7d50\u679c \u767c\u73fe ROUGE-1 \u6700\u9ad8\u60c5\u5f62\u51fa\u73fe\u5728\u5411\uf97e\u9577\ufa01\u70ba 11 \u4e4b\u5167\uff0cROUGE-L \u6700\u9ad8\u51fa\u73fe\u5728\u5411\uf97e\u9577\ufa01\u70ba 9 \u4e4b\u5167\uff0c \u5716 3: \u8abf\u6574\u6982\uf9a3\u5411\uf97e\u9577\ufa01\u8b8a\uf969 \u5716 4: \u8abf\u6574\u5206\u7fa4\uf969\uf97e\u8b8a\uf969 0. 25 0. 27 0. 29 0. 31 0. 33 0. 35 0. 37 0 0. 5 1 1. 5 2 2. 5 3 4 5 5\u7fa4 R OUGE -1 5\u7fa4 R OUGE -L 20\u7fa4 R OUGE -1 20\u7fa4 R OUGE -L score semantic weight \u5f8c\u5206\u5225\u70ba N \u53ca M Source: d30005t\\APW19981104.0772.xml \u5716 2: \u4ee5 attack \u70ba\u4e2d\u5fc3\u4e4b\u8a9e\u610f\u7db2\uf937\u7bc4\uf9b5[12] 3.2. \uf9dd\u7528\u5206\u7fa4\u6280\u8853\u62bd\u53d6\u4e3b\u984c\u6982\uf9a3 \u7279\u5fb5\u8a08\u7b97\u4f4d\u65bc\u540c\u4e00\u6982\uf9a3\u7fa4\u4e2d\u8a9e\uf906\u7684\u91cd\u8981\u6027\u3002 \u9996\u5148\u5c0d\u6b0a\u91cd\u6bd4\uf9b5\u9032\ufa08\u6700\u4f73\u5316\uff0c\u5148\u9078\u64c7\u8a72\u8b8a\uf969\u7684\u539f\u56e0\u662f\u5e0c\u671b\u4e4b\u5f8c\u7684\u5be6\u9a57\u90fd\u53ef\u6709\u4e00\u500b\u6700\u4f73\u7684\u6b0a\u91cd \u5716 5: \u8abf\u6574\u52a0\u5165\u8a9e\u610f\u7db2\uf937\u8b8a\uf969</td></tr><tr><td>\uf906\uff0c\u9054\u5230\u7522\u751f\u7b26\u5408\u539f\u6587\u4e3b\u984c\u6458\u8981\u7684\u76ee\u7684\u3002 \u672c\u6587\u4e2d\u63d0\u51fa\u4ee5\u6982\uf9a3\u5206\u7fa4(Concept Clustering)\u62bd\u53d6\u65b0\u805e\u4e8b\u4ef6\u6240\u63d0\u53ca\u7684\u4e3b\u984c(Topic)\u53ca\u8a9e\u610f\uff0c\u4e26\u7d50 \u5408\u50b3\u7d71\u7279\u5fb5\u9078\u53d6\u6cd5(Feature Selection)\u8a08\u7b97\u8a9e\uf906\u7684\u91cd\u8981\u6027\u53ca\u8a9e\u610f\u6db5\u84cb\ufa01\uff0c\u85c9\u6b64\u4f5c\u70ba\u6311\u9078\u6458\u8981\u8a9e\uf906\u7684 \u6642\u9593\u9806\u5e8f\u3001\u5c08\u6709\u540d\u8a5e\u3001\u5c0d\u4e3b\u984c\u7684\u76f8\u4f3c\ufa01\u4ee5\u53ca\u4ee3\u540d\u8a5e\u7684 Penalty\u3002\u5176\u6311\u9078\u6bb5\uf918\u7684\u4f9d\u64da\u5982\u4e0b\uff1a \u672c\uf941\u6587\u63d0\u51fa\uf978\u7a2e\u6574\u5408\u8a9e\u610f\u7db2\uf937\u65bc\u524d\u5f8c\u6587\u63cf\u8ff0\u7684\u65b9\u6cd5\u3002\u7b2c\u4e00\u7a2e\u65b9\u6cd5\u662f\u5728\u524d\u5f8c\u6587\u4e2d\u53ea\u53d6\u8207\u6982\uf9a3\u65bc \u8a9e\u610f\u7db2\uf937\u4e2d\u6709\u95dc\uf997\u5b57\u8a5e\uff0c\u4ee5\u7f8e\u570b\u5927\u4f7f\u9928(the U.S. Embassy)\u6b64\u4e00\u6982\uf9a3\u70ba\uf9b5\uff0c\u65bc\u8868 1 \u4e2d\uff0c\u63cf\u8ff0\u5b57\u8a5e\u7684 \u96c6\u5408\u5305\u542b Bonn\u3001word\u3001found\u3001officials\u3001Wednesday \u7b49\u4e94\u500b\u5b57\u5f59\uff0c\u7136\u800c\uff0c\u900f\u904e\u8a9e\u610f\u7db2\uf937\u7684\uf99a\u63a5\u5206 \u542b\u6982\uf9a3\uff0c\u7136\u800c\u672c\u6587\u4ea6\u5617\u8a66\u628a\u63cf\u8ff0\u6982\uf9a3\u7684\u5b57\u5f59\u4e5f\u52a0\u9032\u6982\uf9a3\u7fa4\uf9e8\uff1b\u9019\u6a23\u7684\u4f5c\u6cd5\u662f\u5e0c\u671b\u80fd\u5920\u589e\u52a0\u8a9e\uf906\u5c0d \u61c9\u5230\u5b57\u8a5e\u7684\uf969\uf97e\uff0c\u907f\u514d\u4e00\uf906\u8a71\uf9e8\u53ea\u6709\u5c11\uf969\u5e7e\u500b\u5b57\u8a5e\u51fa\u73fe\u5728\u6982\uf9a3\u7fa4\u5167\uff0c\u4e14\u5c0d\u61c9\u7684\u5b57\u8a5e\uf969\uf97e\u8d8a\u591a\uff0c\u8d8a \u4f4d\u65bc\u9996\uf906\u6216\u5c3e\uf906\u7684\u8a9e\uf906\u901a\u5e38\u5177\u6709\u95dc\u9375\u6027\u7684\u8a9e\u610f\u8cc7\u8a0a[5]\u3002\u56e0\u6b64\uff0c\u7576\u8a9e\uf906\u4f4d\u65bc\u6b64\u4f4d\u7f6e\u6642\uff0c\u5247\u52a0 4.3. \u8a9e\uf906\u91cd\u8981\u6027 \u5165\u8a9e\u610f\u7db2\uf937\u53ef\u4ee5\u6709\u6548\u5730\u63d0\u5347\u6458\u8981\u54c1\u8cea\u3002\u7d50\u679c\u4e2d\u4e5f\u986f\u793a\u5206\u7fa4\uf969\u76ee\u5728 5 \u7fa4\u300120 \u7fa4\u6642\u4e92\u6709\u9ad8\u4f4e\uff0c\uf967\u904e \u5716 6 \u6bd4\u8f03\u5728 3.1 \u4e2d\u63d0\u51fa\u7684\uf978\u500b\u52a0\u5165\u8a9e\u610f\u7db2\uf937\u7684\u65b9\u6cd5\uff0c\u65b9\u6cd5\u4e00\u662f\u53ea\u7528\u6709\u51fa\u73fe\u5728\u8a9e\u610f\u7db2\uf937\u4e0a\u7684 \u91cd\u6b64\u8a9e\uf906\u7684\u6b0a\u91cd\u3002 \u6211\u5011\u53ea\u53d6\u6700\u9ad8\u503c\uff0c\u56e0\u6b64\u5728\u9019\u4e00\u7d50\u679c\u4e2d\u6c7a\u5b9a\u5c07\u8a9e\u610f\u7db2\uf937\u52a0\u91cd\u4e4b\u5e38\uf969\u503c X \u70ba 1\uff0c\u5206\u7fa4\uf969\u76ee\u5247\u8a2d\u5b9a\u70ba 5\u3002 \u5b57\u5f59\uf92d\u63cf\u8ff0\u6982\uf9a3\uff0c\u65b9\u6cd5\u4e8c\u662f\u4f7f\u7528\u8a9e\u610f\u7db2\uf937\uf92d\u6c7a\u5b9a\u662f\u5426\u8981\u589e\u52a0\u63cf\u8ff0\u5b57\u5f59\u7684\u6b0a\u91cd\u3002\u53ef\u4ee5\u767c\u73fe\u65b9\u6cd5\u4e8c\u5728 \u5bb9\uf9e0\u5224\u65b7\u8a9e\uf906\u5c6c\u65bc\u54ea\u4e00\u6982\uf9a3\u7fa4\u3002\u7b2c\u4e8c\u500b\u65b9\u6cd5\u5224\u65b7\u8a9e\uf906\u4e2d\u7684\u6982\uf9a3\u5728\u54ea\u500b\u6982\uf9a3\u7fa4\u4e2d\uff0c\u7531\u65bc\u6982\uf9a3\u662f\u4ee5\u7de8 \u6210\u5411\uf97e\u7684\u65b9\u5f0f\u505a K-Means \u5206\u7fa4\uff0c\u6bcf\u500b\u5411\uf97e\u90fd\u53ef\u4ee5\u627e\u51fa\u8207\u6240\u5c6c\u6982\uf9a3\u7fa4\u7684\u76f8\u4f3c\ufa01\uff0c\u4e5f\u5c31\u662f\uf9ea\u4e2d\u5fc3\u9ede \u8a9e\uf906\u8207\u6240\u5c6c\u7684\u6982\uf9a3\u7fa4\u7684\u76f8\u4f3c\ufa01 \u5404\u7a2e\u8b8a\uf969\u7684\u60c5\u6cc1\u4e0b\u90fd\u6bd4\u65b9\u6cd5\u4e00\u8981\u597d\uff0c\u6700\u6975\u7aef\u7684\u60c5\u6cc1\u4e0b\u53ef\u4ee5\u76f8\u5dee 19.6%\u3002\u63a8\u4f30\u539f\u56e0\u6709\u4e8c\uff1a\u7b2c\u4e00\uff0c\u4ee5</td></tr><tr><td>\uf96b\u8003\u4f9d\u64da\u3002\u4ee5\u4e0b\u7c21\u55ae\uf96f\u660e\u672c\u6587\u6240\u63d0\u4e4b\u6458\u8981\u65b9\u6cd5\u7684\uf9ca\u7a0b\uff1a1) \uf9dd\u7528\u524d\u5f8c\u6587\u95dc\u4fc2(Context)\u53ca\u8a9e\u610f\u7db2\uf937 1 MEAD \u63a5\u53d7\u76f8\u95dc\u7684\u6587\u4ef6\u96c6\uff0c\u4ee5\u7522\u751f\u6458\u8981\u3002\u7136\u6b64\u8655\u6240\u63d0\u53ca\u4e4b\u76f8\u95dc\u6587\u4ef6\u96c6\uff0c\u5be6\u70ba\u8003\u616e loosely-related documents\u3002 2 \u6b64\uf9d0\u6587\u7ae0\u901a\u5e38\u65bc\u7b2c\u4e00\u6bb5\u7b2c\u4e00\uf906\uf96f\u660e\u6574\u7bc7\u6587\u7ae0\u7684\u91cd\u9ede\u3002\u56e0\u6b64\uff0c\u9996\uf906\u4e4b\u91cd\u8981\u6027\u5fc5\u9808\u52a0\u91cd\u8003\u616e\u3002 3 \u6982\uf9a3\u70ba\u55ae\u4e00\u6216\u591a\u500b\u5b57\u8a5e\u6240\u7d44\u6210\u7684\u96c6\u5408\u3002\u6b64\u96c6\u5408\u53ef\u8996\u70ba\u4e00\u500b\u6982\uf9a3\u6027\u7684\u63cf\u8ff0\uff0c\u4e26\u5b9a\u7fa9\u8a72\u6982\uf9a3\u7684\u7bc4\u570d\u3002\u900f\u904e\u6b64\u96c6\u5408\uff0c\u53ef\u4f5c\u70ba \u7cfb\u7d71\uf9e4\u89e3\u6982\uf9a3\u8a9e\u610f\u7684\u5a92\u4ecb\u3002 \u7684\u8ddd\uf9ea\u3002\u7576\u8a9e\uf906\uf9e8\u6709\u5411\uf97e\u51fa\u73fe\u5728\u6982\uf9a3\u7fa4\u4e4b\u4e2d\u6642\uff0c\u6703\u4ee5\u8a72\u5411\uf97e\uf9ea\u4e2d\u5fc3\u9ede\u7684\u76f8\u4f3c\ufa01\u7576\u4f5c\u8a72\u8a9e\uf906\u8ddf\u6b64\u6982 3.2 \u7bc0\u63d0\u5230\uf978\u500b\uf967\u540c\u7684\u5c0d\u61c9\u65b9\u5f0f\uff0c\u5206\u5225\u70ba\u6bd4\u8f03\u8a9e\uf906\u8207\u6982\uf9a3\u7fa4\u4e2d\u5171\u540c\u51fa\u73fe\u7684\u5b57\u5f59\uf969\uf97e\u53ca\u6bd4\u8f03\u8a9e</td></tr></table>" | |
| } | |
| } | |
| } | |
| } |