| { | |
| "paper_id": "O13-1019", | |
| "header": { | |
| "generated_with": "S2ORC 1.0.0", | |
| "date_generated": "2023-01-19T08:03:20.650169Z" | |
| }, | |
| "title": "Microblog Sentiment Analysis based on Opinion Target Modifying Relations", | |
| "authors": [ | |
| { | |
| "first": "Jenq-Haur", | |
| "middle": [], | |
| "last": "\u738b\u6b63\u8c6a", | |
| "suffix": "", | |
| "affiliation": { | |
| "laboratory": "", | |
| "institution": "National Taipei University of Technology", | |
| "location": {} | |
| }, | |
| "email": "" | |
| }, | |
| { | |
| "first": "", | |
| "middle": [], | |
| "last": "Wang", | |
| "suffix": "", | |
| "affiliation": { | |
| "laboratory": "", | |
| "institution": "National Taipei University of Technology", | |
| "location": {} | |
| }, | |
| "email": "jhwang@csie.ntut.edu.tw" | |
| }, | |
| { | |
| "first": "Ting-Wei", | |
| "middle": [], | |
| "last": "\u8449\u5ead\u744b", | |
| "suffix": "", | |
| "affiliation": { | |
| "laboratory": "", | |
| "institution": "Taipei University of Technology", | |
| "location": {} | |
| }, | |
| "email": "" | |
| }, | |
| { | |
| "first": "", | |
| "middle": [], | |
| "last": "Ye", | |
| "suffix": "", | |
| "affiliation": { | |
| "laboratory": "", | |
| "institution": "Taipei University of Technology", | |
| "location": {} | |
| }, | |
| "email": "" | |
| } | |
| ], | |
| "year": "", | |
| "venue": null, | |
| "identifiers": {}, | |
| "abstract": "Opinion analysis has grown to be one of the most active research areas in natural language processing. If we can classify reviews and messages of blogs correctly, it will help to analyze product and service competition and to realize the opinion orientations of the people on public issues. In this paper, we propose an opinion orientation estimation approach based on target finding and opinion modifying relations in microblog reviews. First, it collects reviews from microblog and preprocesses the source data. Then, by extracting any entity or aspect of the entity about which an opinion has been expressed according to opinion modifying relations, we calculate the overall score of opinion orientation. In our experiment on the 1000 movie reviews of 50 movies from Twitter, the average accuracy of the proposed method is 84.44%, and the highest precision is 88.89%, which is better than SVM and Naive Bayes. This validates the higher precision from modifying relation identification for opinion orientation classification.", | |
| "pdf_parse": { | |
| "paper_id": "O13-1019", | |
| "_pdf_hash": "", | |
| "abstract": [ | |
| { | |
| "text": "Opinion analysis has grown to be one of the most active research areas in natural language processing. If we can classify reviews and messages of blogs correctly, it will help to analyze product and service competition and to realize the opinion orientations of the people on public issues. In this paper, we propose an opinion orientation estimation approach based on target finding and opinion modifying relations in microblog reviews. First, it collects reviews from microblog and preprocesses the source data. Then, by extracting any entity or aspect of the entity about which an opinion has been expressed according to opinion modifying relations, we calculate the overall score of opinion orientation. In our experiment on the 1000 movie reviews of 50 movies from Twitter, the average accuracy of the proposed method is 84.44%, and the highest precision is 88.89%, which is better than SVM and Naive Bayes. This validates the higher precision from modifying relation identification for opinion orientation classification.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Abstract", | |
| "sec_num": null | |
| } | |
| ], | |
| "body_text": [ | |
| { | |
| "text": "\u5982\"Canon G3 has a great lens.\"\uff0c\u53e5\u5b50\u7d93\u904e\u89e3\u6790\u5f8c\u5f97\u5230 G3 subj\uff0clens obj\uff0cG3 \u70ba\u52d5\u8a5e\"has\"\u7684\u4e3b\u8a5e\uff0clens \u70ba\u52d5\u8a5e\"has\"\u53d7\u8a5e\uff0c\u82e5\u5df2\u77e5\u7684\u76ee\u6a19\u70ba\"lens\"\uff0c\u900f\u904e\u53e5\u5b50\u7684\u7d50\u69cb\u95dc \u4fc2\u8207\u5df2\u77e5\u7684\u76ee\u6a19\u5f97\u77e5\"lens\"\u8207\"G3\"\u70ba\u76f8\u540c\u4e3b\u984c\u7684\u76ee\u6a19\u3002\u4f8b\u5982\"iPod is the best mp3 player\"\uff0c\u53e5\u5b50\u7d93\u904e\u89e3\u6790\u5f8c\u5f97\u5230 iPod subj\uff0cbest \u4fee\u98fe player\uff0c\u800c\"best\"\u70ba\u5df2\u77e5\u7684\u610f\u898b \u8a5e\uff0c\u900f\u904e\u53e5\u5b50\u7684\u7d50\u69cb\u8207\u5df2\u77e5\u7684\u610f\u898b\u8a5e\u5f97\u77e5\"player\"\u8207\"iPod\"\u70ba\u76f8\u540c\u4e3b\u984c\u7684\u76ee\u6a19\u3002 (\u4e8c) \u3001\u4fee\u98fe\u95dc\u4fc2\u8fa8\u8b58 to death.\uff0c\"film\"\u662f\u6211\u5011\u627e\u5230\u7684 target\uff0c\"bored\"\u662f\u53e5\u5b50\u4e2d\u7684 VB\uff0c\u56e0\u70ba\" bored\"\u662f\u5c6c\u65bc \u8ca0\u9762\u60c5\u7dd2\u8a5e\uff0c\u6240\u4ee5\u6b64\u53e5\u662f\u5c6c\u65bc\u5c0d\u96fb\u5f71\u7684\u8ca0\u9762\u7684\u8a55\u50f9\u3002 3. T + VB/ VBD/ VBG/ VBN/ VBP/ VBZ + JJ (\u610f\u898b\u8a5e\uff0c\u5f62\u5bb9\u8a5e): T \u70ba target\uff0c\u662f\u53e5\u5b50\u4e2d \u7684\u4e3b\u8a5e\u3002\u4f8b\u5982: This movie is worth seeing.\uff0c\"movie\"\u662f\u6211\u5011\u627e\u5230\u7684 target\uff0c\"is\"\u662f\u53e5\u5b50 \u4e2d\u7684 VBZ\uff0c\"worth\"\u5728\u53e5\u5b50\u4e2d\u70ba JJ\uff0c\u56e0\u70ba\" worth\"\u662f\u5c6c\u65bc\u6b63\u9762\u60c5\u7dd2\u8a5e\uff0c\u6240\u4ee5\u6b64\u53e5\u662f\u5c6c Curtis's acting is amazing.\uff0c\u4f8b\u5b50\u4e2d\u627e\u5230\u7b2c 3 \u7a2e\u7279\u5fb5 T + VBZ + JJ\uff0c\u4f46\u5728 target finding \u6642 \u6211\u5011\u627e\u5230\"movie\"\u53ca\"acting\"\u5169\u500b target\uff0c\"is\"\u662f VBZ\uff0c\u9019\u6642\u6703\u5c0b\u627e\u8207\u4fee\u98fe\u8a5e JJ \"amazing\" \u6700\u8fd1\u7684\u5b57\uff0c\u4e5f\u5c31\u662f\"acting\"\uff0c\u6700\u5f8c\u627e\u5230\u7684\u7279\u5fb5\u5c31\u662f\"acting + is + amazing\"\u3002\u5728\u6b64\u7279\u5fb5\u4e2d\u6703", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "\u524d \u7f6e \u8655 \u7406 \uff0c \u5982 : \u53e5 \u5b50 \u7c21 \u5316 \uff0c \u53e6 \u5916 \u6839 \u64da \u67e5 \u8a62 \u4e3b \u984c \u7684 \u4e0d \u540c \uff0c \u6211 \u5011 \u5229 \u7528 \u4e3b \u984c \u76f8 \u95dc \u8cc7 \u6e90 (topic-", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "(2) \u3001\u6bd4\u8f03\u53e5 \u6240\u8b02\u300c\u6bd4\u8f03\u7d1a\u300d\u5c31\u662f\u5728\u96d9\u65b9\u6216\u5169\u8005\u9593\u505a\u6bd4\u8f03\u7684\u8868\u9054\u65b9\u5f0f\uff0c\u6bd4\u8f03\u7684\u5167\u5bb9\u7576\u7136\u5c31\u4e0d\u5916\u662f \u300c\u5f62\u5bb9\u8a5e\u300d\u6216\u300c\u526f\u8a5e\u300d\u4e86\uff0c\u4f7f\u7528\u8005\u5728\u8a55\u8ad6\u67d0\u4e00\u7269\u4ef6\u6642\u5e38\u6703\u4ee5\u76f8\u4f3c\u7684\u7269\u4ef6\u4f86\u505a\u6bd4\u8f03\uff0c\u4f8b\u5982 :", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "\u56e0\u70ba\u5b57\u8207\u5b57\u5728\u53e5\u5b50\u4e2d\u7684\u8ddd\u96e2\u800c\u5f71\u97ff\u6b63\u8ca0\u9762\u8a55\u50f9\u7684\u5206\u6578\uff0c\u6211\u5011\u5c07\u5728\u5f8c\u9762\u7ae0\u7bc0\u4f5c\u4ecb\u7d39\u3002", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "The picture quality of Camera-x is better than that of Camera-y.\u5728\u4f8b\u5b50\u4e2d\u6bd4\u8f03\u5169\u53f0\u76f8\u6a5f\u7684\u76f8 \u7247\u756b\u8cea\uff0c\u9019\u662f\u6700\u5e38\u898b\u7684\u6bd4\u8f03\u95dc\u4fc2\u3002\u6211\u5011\u5c07\u5e38\u898b\u7684\u6bd4\u8f03\u95dc\u4fc2\u5206\u70ba\u4e0b\u5217\u5169\u9805 [14] [15]:", | |
| "cite_spans": [ | |
| { | |
| "start": 109, | |
| "end": 113, | |
| "text": "[14]", | |
| "ref_id": "BIBREF13" | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "\u56e0\u70ba\u5b57\u8207\u5b57\u5728\u53e5\u5b50\u4e2d\u7684\u8ddd\u96e2\u800c\u5f71\u97ff\u6b63\u8ca0\u9762\u8a55\u50f9\u7684\u5206\u6578\uff0c\u6211\u5011\u5c07\u5728\u5f8c\u9762\u7ae0\u7bc0\u4f5c\u4ecb\u7d39\u3002", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "1. \u975e\u5c0d\u7b49\u6bd4\u8f03 (non-equal comparisons) : \u7269\u4ef6\u9593\u6bd4\u8f03\u5c6c\u6027\u7684\u512a\u52a3\uff0c\u4f8b\u5982 : The VIA chip is faster than that of AMD.\uff0c\u4f8b\u5b50\u4e2d\u662f\u6700\u5e38\u898b\u7684\u6bd4\u8f03\u7269\u4ef6\u5c6c\u6027\u7684\u512a\u52a3\u95dc\u4fc2\u3002\u53c8\u4f8b\u5982 : I prefer VIA to AMD.\uff0c\u6b64\u4f8b\u5b50\u4e5f\u662f\u8868\u9054\u512a\u52a3\u95dc\u4fc2\u3002\u6211\u5011\u5229\u7528 POS tagger \u4f86\u6a19\u8a18\u6bd4\u8f03 \u7d1a\uff0c\u6a19\u8a18\u70ba\"JJR\"\u3001\"JJS\"\u3001\"RBR\"\u3001\"RBS\"\uff0c\u70ba\u5f62\u5bb9\u8a5e\u53ca\u526f\u8a5e\u7684\u6bd4\u8f03\u7d1a\uff0c\u4f8b\u5982: Life was harder then because neither of us had a job.\uff0c\u4f8b\u5b50\u4e2d\"harder\"\u7d93\u904e POS tagger \u6a19\u8a3b \u70ba\"JJR\"\u3002\u518d\u4f86\u5c0b\u627e target \u8207\u6a19\u8a18\u7684\u76f8\u5c0d\u4f4d\u7f6e: ", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "\u56e0\u70ba\u5b57\u8207\u5b57\u5728\u53e5\u5b50\u4e2d\u7684\u8ddd\u96e2\u800c\u5f71\u97ff\u6b63\u8ca0\u9762\u8a55\u50f9\u7684\u5206\u6578\uff0c\u6211\u5011\u5c07\u5728\u5f8c\u9762\u7ae0\u7bc0\u4f5c\u4ecb\u7d39\u3002", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "\uf0d8 Target +", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "\u56e0\u70ba\u5b57\u8207\u5b57\u5728\u53e5\u5b50\u4e2d\u7684\u8ddd\u96e2\u800c\u5f71\u97ff\u6b63\u8ca0\u9762\u8a55\u50f9\u7684\u5206\u6578\uff0c\u6211\u5011\u5c07\u5728\u5f8c\u9762\u7ae0\u7bc0\u4f5c\u4ecb\u7d39\u3002", | |
| "sec_num": null | |
| } | |
| ], | |
| "back_matter": [], | |
| "bib_entries": { | |
| "BIBREF0": { | |
| "ref_id": "b0", | |
| "title": "Annotation and Detection of Emotion in a Task Oriented Human-Human Dialog Corpus", | |
| "authors": [ | |
| { | |
| "first": "L", | |
| "middle": [], | |
| "last": "Devillers", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "I", | |
| "middle": [], | |
| "last": "Vasilescu", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "L", | |
| "middle": [], | |
| "last": "Lamel", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2002, | |
| "venue": "Proceedings of the ISLE Workshop on Dialogue Tagging for Multi-Modal Human-Computer Interaction", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "624--629", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Devillers L., Vasilescu I., and Lamel L., \"Annotation and Detection of Emotion in a Task Oriented Human-Human Dialog Corpus,\" Proceedings of the ISLE Workshop on Dialogue Tagging for Multi-Modal Human-Computer Interaction, pp.624-629, 2002.", | |
| "links": null | |
| }, | |
| "BIBREF1": { | |
| "ref_id": "b1", | |
| "title": "Prosodic Cues for Emotion Recognition in Communicator Dialogs", | |
| "authors": [ | |
| { | |
| "first": "J", | |
| "middle": [ | |
| "C" | |
| ], | |
| "last": "Ang", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2002, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Ang J. C., \"Prosodic Cues for Emotion Recognition in Communicator Dialogs,\" M.S. thesis, University of California Berkeley, 2002.", | |
| "links": null | |
| }, | |
| "BIBREF2": { | |
| "ref_id": "b2", | |
| "title": "Extracting opinions, opinion holders, and topics expressed in online news media text", | |
| "authors": [ | |
| { | |
| "first": "S", | |
| "middle": [], | |
| "last": "Kim", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "E", | |
| "middle": [], | |
| "last": "Hovy", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2006, | |
| "venue": "Proceedings of ACL/COLING Workshop on Sentiment and Subjectivity in Text", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "1--8", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "S. Kim and E. Hovy. \"Extracting opinions, opinion holders, and topics expressed in online news media text,\" In Proceedings of ACL/COLING Workshop on Sentiment and Subjectivity in Text, pp.1-8 ,2006.", | |
| "links": null | |
| }, | |
| "BIBREF3": { | |
| "ref_id": "b3", | |
| "title": "Extracting product features and opinions from reviews", | |
| "authors": [ | |
| { | |
| "first": "A", | |
| "middle": [ | |
| "M" | |
| ], | |
| "last": "Popescu", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "O", | |
| "middle": [], | |
| "last": "Etzioni", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2005, | |
| "venue": "Proceedings of the conference on Human Language Technology and Empirical Methods in Natural Language Processing", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "339--346", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Popescu A. M., Etzioni O., \"Extracting product features and opinions from reviews,\" Proceedings of the conference on Human Language Technology and Empirical Methods in Natural Language Processing, p.339-346, October 06-08, Vancouver, British Columbia, Canada, 2005.", | |
| "links": null | |
| }, | |
| "BIBREF4": { | |
| "ref_id": "b4", | |
| "title": "Opinion word expansion and target extraction through double propagation", | |
| "authors": [ | |
| { | |
| "first": "G", | |
| "middle": [], | |
| "last": "Qiu", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "B", | |
| "middle": [], | |
| "last": "Liu", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "J", | |
| "middle": [], | |
| "last": "Bu", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Chen", | |
| "middle": [ | |
| "C" | |
| ], | |
| "last": "", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2011, | |
| "venue": "Computational Linguistics", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "363--370", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Qiu, G., Liu B., Bu J., and Chen C.. \"Opinion word expansion and target extraction through double propagation,\" Computational Linguistics, pp.363-370, 2011.", | |
| "links": null | |
| }, | |
| "BIBREF5": { | |
| "ref_id": "b5", | |
| "title": "Movie review mining and summarization", | |
| "authors": [ | |
| { | |
| "first": "L", | |
| "middle": [], | |
| "last": "Zhuang", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "F", | |
| "middle": [], | |
| "last": "Jing", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "X.-Y", | |
| "middle": [], | |
| "last": "Zhu", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2006, | |
| "venue": "Proceedings of the 15th ACM international conference on Information and knowledge management", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "43--50", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Zhuang L., Jing F., Zhu X.-Y., \"Movie review mining and summarization,\" Proceedings of the 15th ACM international conference on Information and knowledge management, Arlington, Virginia, USA, pp. 43-50, November 06-11, 2006", | |
| "links": null | |
| }, | |
| "BIBREF6": { | |
| "ref_id": "b6", | |
| "title": "Twitter Sentiment Classification using Distant Supervision", | |
| "authors": [ | |
| { | |
| "first": "A", | |
| "middle": [], | |
| "last": "Go", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "R", | |
| "middle": [], | |
| "last": "Bhayani", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "L", | |
| "middle": [], | |
| "last": "Huang", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": null, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Go A., Bhayani R., and Huang L., \"Twitter Sentiment Classification using Distant Supervision,\" Technical report, Stanford Digital Library Technologies Project.", | |
| "links": null | |
| }, | |
| "BIBREF7": { | |
| "ref_id": "b7", | |
| "title": "Twitter as a corpus for sentiment analysis and opinion mining", | |
| "authors": [ | |
| { | |
| "first": "A", | |
| "middle": [], | |
| "last": "Pak", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "P", | |
| "middle": [], | |
| "last": "Paroubek", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2010, | |
| "venue": "Proceedings of LREC", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "1320--1326", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Pak A. and Paroubek P., \"Twitter as a corpus for sentiment analysis and opinion mining,\" Proceedings of LREC, pp.1320-1326, 2010.", | |
| "links": null | |
| }, | |
| "BIBREF8": { | |
| "ref_id": "b8", | |
| "title": "Sentiment Classification of Short Chinese Sentences", | |
| "authors": [ | |
| { | |
| "first": "Y", | |
| "middle": [ | |
| "T" | |
| ], | |
| "last": "Sun", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "C", | |
| "middle": [ | |
| "L" | |
| ], | |
| "last": "Chen", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "C", | |
| "middle": [ | |
| "C" | |
| ], | |
| "last": "Liu", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "C", | |
| "middle": [ | |
| "L" | |
| ], | |
| "last": "Liu", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "V", | |
| "middle": [ | |
| "W" | |
| ], | |
| "last": "Soo", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2010, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Sun Y. T., Chen C. L., Liu C. C., Liu C. L., Soo V. W., \"Sentiment Classification of Short Chinese Sentences,\" ROCLING 2010.", | |
| "links": null | |
| }, | |
| "BIBREF9": { | |
| "ref_id": "b9", | |
| "title": "Emotion Modeling from Writer/Reader Perspective Using a Microblog Dataset", | |
| "authors": [ | |
| { | |
| "first": "Y", | |
| "middle": [], | |
| "last": "Tang", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "H", | |
| "middle": [], | |
| "last": "Chen", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2011, | |
| "venue": "Workshop on Sentiment Analysis where AI meets Psychology (SAAIP 2011)", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "11--19", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Tang, Y., Chen, H. \"Emotion Modeling from Writer/Reader Perspective Using a Microblog Dataset,\" In: Workshop on Sentiment Analysis where AI meets Psychology (SAAIP 2011), pp. 11-19. ACL, 2011.", | |
| "links": null | |
| }, | |
| "BIBREF10": { | |
| "ref_id": "b10", | |
| "title": "Enhanced sentiment learning using twitter hashtags and smileys", | |
| "authors": [ | |
| { | |
| "first": "D", | |
| "middle": [], | |
| "last": "Davidov", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "O", | |
| "middle": [], | |
| "last": "Tsur", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "A", | |
| "middle": [], | |
| "last": "Rappoport", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2010, | |
| "venue": "Proceedings of Coling", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "241--249", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Davidov, D.; Tsur, O.; and Rappoport, A. \"Enhanced sentiment learning using twitter hashtags and smileys,\" In Proceedings of Coling, pp.241-249 ,2010.", | |
| "links": null | |
| }, | |
| "BIBREF11": { | |
| "ref_id": "b11", | |
| "title": "Creating subjective and objective sentence classifiers from unannotated texts", | |
| "authors": [ | |
| { | |
| "first": "J", | |
| "middle": [], | |
| "last": "Wiebe", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "E", | |
| "middle": [], | |
| "last": "Riloff", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2005, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "486--497", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Wiebe J., Riloff E.. \"Creating subjective and objective sentence classifiers from unannotated texts,\" pp.486-497 , In CICLing-2005.", | |
| "links": null | |
| }, | |
| "BIBREF12": { | |
| "ref_id": "b12", | |
| "title": "Robust sentiment detection on twitter from biased and noisy data", | |
| "authors": [ | |
| { | |
| "first": "L", | |
| "middle": [], | |
| "last": "Barbosa", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "J", | |
| "middle": [], | |
| "last": "Feng", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2010, | |
| "venue": "Proc. of Coling", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "36--44", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Barbosa, L., Feng, J. \"Robust sentiment detection on twitter from biased and noisy data,\" In Proc. of Coling, pp.36-44, 2010.", | |
| "links": null | |
| }, | |
| "BIBREF13": { | |
| "ref_id": "b13", | |
| "title": "Mining comparative sentences and relations", | |
| "authors": [ | |
| { | |
| "first": "N", | |
| "middle": [], | |
| "last": "Jindal", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "B", | |
| "middle": [], | |
| "last": "Liu", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2006, | |
| "venue": "AAAI'06", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Jindal, N., and Liu, B. \"Mining comparative sentences and relations,\" AAAI'06, 2006.", | |
| "links": null | |
| }, | |
| "BIBREF14": { | |
| "ref_id": "b14", | |
| "title": "Identifying comparative sentences in text documents", | |
| "authors": [ | |
| { | |
| "first": "N", | |
| "middle": [], | |
| "last": "Jindal", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "B", | |
| "middle": [], | |
| "last": "Liu", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2006, | |
| "venue": "SIGIR-06", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Jindal, N. and Liu, B. \"Identifying comparative sentences in text documents,\" SIGIR-06, 2006.", | |
| "links": null | |
| }, | |
| "BIBREF15": { | |
| "ref_id": "b15", | |
| "title": "Negation in English Speech and Writing: A Study in Variation", | |
| "authors": [ | |
| { | |
| "first": "", | |
| "middle": [], | |
| "last": "Tottie", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "", | |
| "middle": [], | |
| "last": "Gunnel", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1991, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Tottie. Gunnel. \"Negation in English Speech and Writing: A Study in Variation,\" San Diego: Academic Press, 1991.", | |
| "links": null | |
| }, | |
| "BIBREF16": { | |
| "ref_id": "b16", | |
| "title": "Negation in Natural Language: On the Form and Meaning of Negative Elements", | |
| "authors": [ | |
| { | |
| "first": "H", | |
| "middle": [], | |
| "last": "Zeijlstra", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2007, | |
| "venue": "Language and Linguistics Compass", | |
| "volume": "1", | |
| "issue": "5", | |
| "pages": "498--518", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "H. Zeijlstra. \"Negation in Natural Language: On the Form and Meaning of Negative Elements,\" Language and Linguistics Compass, 1(5):498-518, 2007.", | |
| "links": null | |
| }, | |
| "BIBREF17": { | |
| "ref_id": "b17", | |
| "title": "Opinion mining and sentiment analysis", | |
| "authors": [ | |
| { | |
| "first": "B", | |
| "middle": [], | |
| "last": "Pang", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "L", | |
| "middle": [], | |
| "last": "Lee", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2008, | |
| "venue": "Foundations and Trends in Information Retrieval", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "B. Pang and L. Lee. \"Opinion mining and sentiment analysis,\" Foundations and Trends in Information Retrieval, 2008.", | |
| "links": null | |
| }, | |
| "BIBREF18": { | |
| "ref_id": "b18", | |
| "title": "Machine Learning", | |
| "authors": [ | |
| { | |
| "first": "Tom", | |
| "middle": [ | |
| "M" | |
| ], | |
| "last": "Mitchell", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1997, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Tom M. Mitchell, Machine Learning, WCB-McGraw-Hill, 1997", | |
| "links": null | |
| }, | |
| "BIBREF20": { | |
| "ref_id": "b20", | |
| "title": "Minipar", | |
| "authors": [], | |
| "year": null, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Minipar, http://webdocs.cs.ualberta.ca/~lindek/minipar.htm", | |
| "links": null | |
| }, | |
| "BIBREF21": { | |
| "ref_id": "b21", | |
| "title": "The Stanford Parser: A statistical parser", | |
| "authors": [], | |
| "year": null, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "The Stanford Parser: A statistical parser, http://nlp.stanford.edu/software/lex-parser.shtml [23] Dependency grammar, http://en.wikipedia.org/wiki/Dependency_grammar", | |
| "links": null | |
| } | |
| }, | |
| "ref_entries": { | |
| "FIGREF0": { | |
| "type_str": "figure", | |
| "text": "JJ (\u610f\u898b\u8a5e\uff0c\u5f62\u5bb9\u8a5e) + T: \u6b64\u9805\u898f\u5247\u4e3b\u8981\u662f\u627e\u5230 target \u53ca\u4fee\u98fe target \u7684\u4fee\u98fe\u8a5e\u3002\u4f8b\u5982:It's my favorite movie.\uff0c\u4f8b\u5b50\u4e2d\uff0c\"movie\"\u662f target\uff0c\"favorite\"\u662f JJ\uff0c\u4e5f\u662f\u4fee\u98fe target \u7684\u5f62 \u5bb9\u8a5e\uff0c\u56e0\u70ba\" favorite\"\u662f\u5c6c\u65bc\u6b63\u9762\u60c5\u7dd2\u8a5e\uff0c\u6240\u4ee5\u6b64\u53e5\u662f\u5c6c\u65bc\u5c0d\u96fb\u5f71\u7684\u6b63\u9762\u7684\u8a55\u50f9\u3002 \u4ee5\u4e0a\u7684\u4fee\u98fe\u95dc\u4fc2\uff0c\u90fd\u662f\u5c0b\u627e\u53e5\u5b50\u4e2d\u8ddd\u96e2\u6700\u8fd1\u7684\u55ae\u5b57\uff0c\u4f8b\u5982: In the first movie Tony", | |
| "num": null, | |
| "uris": null | |
| }, | |
| "TABREF0": { | |
| "html": null, | |
| "text": "\u4e00\u3001\u7dd2\u8ad6 \u60c5\u7dd2\u5075\u6e2c (emotion detection)\u7684\u767c\u5c55\u5c0d\u65bc\u5546\u696d\u8207\u79d1\u6280\u7684\u4e92\u52d5\u5177\u6709\u9ad8\ufa01\u7684\u61c9\u7528\u50f9\u503c\uff0c Hovy[3]\u91dd\u5c0d\u5404\u985e\u4e3b\u984c\u7684\u65b0\u805e\u9032\u884c\u627e\u51fa\u5167\u6587\u4e2d\u7684 opinion holders \u53ca opinion topics\u3002\u9996\u5148\u4ee5\u52d5\u8a5e\u53ca\u5f62\u5bb9\u8a5e\u70ba\u4e3b\u5efa\u7acb\u60c5\u7dd2\u8fad\u5178\uff0c\u63a5\u8457\u7136\u5f8c\u4f7f\u7528\u5256\u6790\u5668\u89e3\u6790\u53e5\u5b50\uff0c\u4e26\u5c07 FrameNet \u7684 frame element \u53ca\u7bc4\u4f8b\u53e5\u5b50\u4f86\u9032\u884c Maximum Entropy \u8a13\u7df4\u4ee5\u627e\u5230\u53e5\u5b50\u4e2d\u7684 opinion holder \u53ca opinion topic\u3002\u6700\u5f8c\u5be6\u9a57\u7d50\u679c\u7684\u6e96\u78ba\u5ea6\u70ba 64%\uff0c\u8aaa\u660e FrameNet \u4e2d\u7684 frame \u53ca frame element \u6709\u9650\uff0c\u53ea\u80fd\u627e\u5230\u90e8\u5206\u7684\u53e5\u5b50\u7d50\u69cb\uff0c\u56e0\u6b64\u6e96\u78ba\u7387\u4e26\u4e0d\u9ad8\u3002 Popescu \u548c Etzioni[4] \u91dd \u5c0d \u5546 \u54c1 \u7684 \u4f7f \u7528 \u8005 \u8a55 \u8ad6 \uff0c \u63a1 \u7528 PMI (Point-wise Mutual Information) \u4f86\u7372\u5f97\u8207\u4e3b\u984c\u5171\u540c\u51fa\u73fe\u6a5f\u7387\u6700\u9ad8\u7684\u8a5e\u4f86\u7576\u4f5c\u8a55\u8ad6\u76ee\u6a19 (opinion targets)\uff0c\u8207 \u672c\u7814\u7a76\u5206\u6cd5\u985e\u4f3c\u3002\u6211\u5011\u9664\u4e86\u4f7f\u7528 PMI \u4ee5\u5916\uff0c\u6709\u9451\u65bc\u4e0d\u540c\u4f7f\u7528\u8005\u6240\u4f7f\u7528\u7684\u5b57\u8a5e\u6703\u6709\u6240\u4e0d \u540c\uff0c\u6240\u4ee5\u4e5f\u5c07 PMI \u6240\u7372\u5f97\u8a5e\u7684\u540c\u7fa9\u5b57\u4f86\u64f4\u589e\u6211\u5011\u7684\u8a55\u8ad6\u76ee\u6a19\u3002 Qiu \u7b49\u4eba[5]\u4f7f\u7528\u6d88\u8cbb\u8005\u8a55\u8ad6\u8cc7\u6599\u96c6[20]\u88e1\u7684\u8a55\u8ad6\u8fa8\u8b58\u51fa\u8a55\u8ad6\u76ee\u6a19(target)\u53ca\u610f\u898b\u8a5e (opinion word)\u3002\u9996\u5148\u4f7f\u7528 POS tagging \u6a19\u8a3b\u5b57\u7684\u8a5e\u6027\uff0c\u4ed6\u5011\u5b9a\u7fa9 target \u70ba\u540d\u8a5e\uff0copinion words \u70ba\u5f62\u5bb9\u8a5e\u3002\u518d\u900f\u904e Minipar[21]\u89e3\u6790\u53e5\u5b50\u7684\u7d50\u69cb\u3002\u6700\u5f8c\u5229\u7528\u53e5\u5b50\u7684\u7d50\u69cb\u53ca\u8cc7\u6599\u5df2\u6a19 \u8a18\u597d\u7684\u76ee\u6a19 (\u5546\u54c1\u76f8\u95dc\u5c6c\u6027\u7b49\u7b49)\u548c \u610f\u898b\u8a5e\u4f86\u627e\u5230\u53e5\u5b50\u4e2d\u672a\u77e5\u800c\u53ef\u80fd\u7684\u76ee\u6a19\u3002\u4f8b", | |
| "content": "<table><tr><td>\u5305\u62ec\u4f9d\u7167\u4f7f\u7528\u8005\u60c5\u7dd2\u63a8\u85a6\u76f8\u7b26\u5408\u7684\u6587\u7ae0\u3001\u97f3\uf914\u7b49\u5546\u54c1\u3002\u672c\u7814\u7a76\u900f\u904e\u77e5\u540d\u5fae\u7db2\u8a8c Twitter</td></tr><tr><td>\u7684\u82f1\u6587\u77ed\uf906\u4e2d\u7684\u60c5\u7dd2\u8a5e\u5f59\u9032\ufa08\u63a8\u6587 (tweet) \u60c5\u7dd2\u5206\uf9d0\uff0c\u56e0\u70ba\u77ed\u7bc7\u6587\u4ef6\u6240\u5305\u542b\u7684\u8a9e\u5883\u548c\u8a5e</td></tr><tr><td>\u5f59\u901a\u5e38\u6bd4\u8f03\uf967\u8db3\u5920\uff0c\u6240\u4ee5\u77ed\u7bc7\u6587\u4ef6\u7684\u6587\u4ef6\u5206\uf9d0\u6548\u679c\u901a\u5e38\u6703\u6bd4\u9577\u7bc7\u7684\u6587\u4ef6\u5206\uf9d0\u6548\u679c\uf967\u4f73\u3002</td></tr><tr><td>\u6709\u5225\u65bc\u50b3\u7d71\u6587\u4ef6\u5206\uf9d0\uff0c\u6211\u5011\u5206\u6790\u60c5\u7dd2\u8a5e\u5f59\u8207\u4fee\u98fe\u95dc\u4fc2\u9032\u884c\u4ee5\u53e5\u5b50\u70ba\u57fa\u790e\u7684\u60c5\u7dd2\u5075\u6e2c</td></tr><tr><td>(sentence-based emotion detection) \u554f\u984c\u3002</td></tr><tr><td>\u672c\u7814\u7a76\u65b9\u6cd5\u4f7f\u7528\u82f1\u6587\u6587\u6cd5\u7684\u4fee\u98fe\u95dc\u4fc2\uff0c\u4e3b\u8981\u662f tweet \u4e2d\u7684\u5167\u5bb9\u8a55\u8ad6\u76ee\u6a19\u8207\u610f\u898b\u8a5e\u4e4b</td></tr><tr><td>\u9593\u7684\u4fee\u98fe\u95dc\u4fc2\uff0c\u627e\u5230\u4fee\u98fe\u95dc\u4fc2\u5373\u80fd\u5224\u65b7\u8a55\u8ad6\u8005\u85c9\u6b64\u5167\u5bb9\u6292\u767c\u67d0\u7a2e\u60c5\u7dd2\u3002\u6839\u64da\u8a55\u8ad6\u4e3b\u984c\u4ee5</td></tr><tr><td>\u53ca\u610f\u898b\u8a5e\u7684\u4fee\u98fe\u95dc\u4fc2\uff0c\u767c\u6398\u51fa\u4e3b\u984c\u76f8\u95dc\u7684\u8a55\u8ad6\u76ee\u6a19\u4ee5\u5224\u65b7\u5176\u610f\u898b\u50be\u5411\u4f86\u9810\u6e2c\u672a\u77e5\u60c5\u7dd2\uf9d0</td></tr><tr><td>\u5225\u7684\u6587\u7ae0\u4e4b\u53ef\u80fd\u60c5\u7dd2\u3002</td></tr><tr><td>\u4e8c\u3001\u76f8\u95dc\u7814\u7a76</td></tr><tr><td>\u5e38\u898b\u7684\u60c5\u7dd2\u5075\u6e2c\u65b9\u6cd5\u6240\u9069\u7528\u7684\u7bc4\u570d\u53ef\u5206\u70ba\u70ba\uf906\u5b50\u5c64\u6b21\u7684\u63a8\uf941\uff0c\u6bb5\uf918\u5c64\u6b21\u548c\u5168\u7bc7\u6587\u7ae0</td></tr><tr><td>\u5c64\u6b21\u7684\u60c5\u7dd2\u5075\u6e2c\u65b9\u6cd5 [1][2]\u3002\u56e0\u70ba\u5fae\u7db2\u8a8c\u7684\u5b57\u6578\u9650\u5236\uff0c\u672c\u7bc7\u7814\u7a76\u5c08\u6ce8\u5728\uf906\u5b50\u5c64\u6b21\u7684\u60c5\u7dd2</td></tr><tr><td>\u5075\u6e2c\u65b9\u6cd5\uff0c\u5305\u62ec\u8a55\u8ad6\u76ee\u6a19 (target)\u3001\u610f\u898b\u8a5e (opinion word)\u7b49\u3002\u63a5\u4e0b\uf92d\u5c07\u63a2\u8a0e\u4e00\u4e9b\u4f7f\u7528\u6587</td></tr><tr><td>\u4ef6\u5206\uf9d0\u76f8\u95dc\u6280\u8853\u65bc\u60c5\u7dd2\u5075\u6e2c\u7684\u6587\u737b\uff0c\u5f7c\u6b64\u6700\u5927\u7684\u5dee\uf962\u5728\u65bc\u5075\u6e2c\u65b9\u6cd5\u4e0a\u7684\u5dee\uf962\u3002</td></tr><tr><td>(\u4e00) \u3001\u8a55\u8ad6\u76ee\u6a19\u767c\u6398</td></tr><tr><td>Kim \u548c</td></tr></table>", | |
| "type_str": "table", | |
| "num": null | |
| }, | |
| "TABREF2": { | |
| "html": null, | |
| "text": "After seeing Prometheus, I was hoping everyone had forgotten about it. http://www.deadline.com/hollywood.\"\u3002 \uf0d8 Retweet (RT): \u5c31\u662f\u8f49\u63a8\u7684\u610f\u601d\uff0c\u7576\u4f60\u5728Twitter\u4e0a\u770b\u5230\u4e00\u500b\u6709\u610f\u601d\u7684tweet\uff0c\u5c31\u53ef\u4ee5 RT\u4e00\u4e0b\uff0c\u4ee5\u5e6b\u52a9\u50b3\u64ad\u9019\u689d\u4fe1\u606f\u3002\u7528\u6cd5\u70ba\uff1aRT @\u539f\u59cb\u767c\u5e03\u8005Twitter ID: \u88ab\u8f49\u63a8\u7684\u539f \u6587\u3002\u4f8b\u5982 : \"RT if you like Titanic, Harry Potter, Twilight, Pitch Perfect, Skyfall, Life of Pi, Transformers, Les Miserables & etc.. :)\"\u3002 went to theater to watch Argo yesterday. Ring Ring Ring! I was so humiliated when my phone rang out.\uff0c\u4f8b\u5b50\u4e2d\u5c6c\u65bc\u8ca0\u9762\u7684\u60c5\u7dd2\u6292\u767c\uff0c\u4f46\u8a55\u8ad6\u7684\u662f\u56e0\u70ba\u96fb\u5f71 \u6211\u5011\u5728 Movie Review Data \u6587\u96c6[30]\u4f7f\u7528 PMI (Point-wise Mutual Information)\u8655\u7406\u8a5e \u5f59\u5171\u540c\u51fa\u73fe\u95dc\u4fc2 (word collocation)\uff0c\u9032\u800c\u4e86\u89e3\u6587\u96c6\u4e2d\u4f7f\u7528\u8005\u8a55\u8ad6\u96fb\u5f71\u6642\u6700\u5e38\u63d0\u53ca\u7684\u540d \u8a5e\u3002\u6211\u5011\u5229\u7528\u6b64\u6587\u96c6\u6240\u6709\u7684\u540d\u8a5e\u55ae\u7368\u51fa\u73fe\u6b21\u6578 (term frequency, tf) \u548c\u8207\u55ae\u5b57\"movie\" \u5171 \u540c\u51fa\u73fe\u7684\u8a5e\u5f59\u7d44\u5408 (emotion-words collocation pairs)\u51fa\u73fe\u6b21\u6578\uff0c\u5206\u5225\u8a08\u7b97 PMI score \u4e26\u4f9d \u7167\ufa09\u51aa\u6392\uf99c\uff0c\u518d\u5f9e\u4e2d\u53d6 PMI \u6700\u9ad8\u7684\u524d k \u500b\u8a5e\u5f59\u5171\u540c\u51fa\u73fe\u7684\u7d44\u5408\u4f5c\u70ba\u7279\u5fb5\u5b50\u96c6\u5408\u3002\u6700\u5f8c \u5728\u5be6\u9a57\u4e2d\u8a55\u4f30 k \u61c9\u8a72\u53d6\u5e7e\u500b\u5b57\u4f86\u7576\u4f5c\u6211\u5011\u7684 target\u3002", | |
| "content": "<table><tr><td>1\u3001 \u62fc\u5b57\u6aa2\u67e5 1. \u8a55\u8ad6\u76ee\u6a19\u767c\u6398 2. \u8a5e\u6027\u4fee\u98fe\u95dc\u4fc2</td><td/></tr><tr><td colspan=\"2\">\u5728 Twitter \u4e2d\uff0c\u4f7f\u7528\u8005\u5f80\u5f80\u4e0d\u6703\u592a\u7559\u610f\u62fc\u5b57\u7684\u6b63\u78ba\u6027\uff0c\u6709\u6642\u4e5f\u6703\u85c9\u7531\u55ae\u5b57\u4f86\u5f37\u8abf\u4ed6 \u96d6\u7136\u6211\u5011\u5f9eTwitter\u4e2d\u4f9d\u7167\u4e3b\u984c\u6536\u96c6tweet\uff0c\u4f46tweet\u5167\u5bb9\u4ecd\u53ef\u80fd\u5305\u542b\u767c\u6587\u8005\u5c0d\u7121\u95dc\u4e3b \u5728\u627e\u5230\u53ef\u80fd\u7684\u8a55\u8ad6\u76ee\u6a19 (target)\u5f8c\uff0c\u63a5\u4e0b\u4f86\u8981\u627e\u5230\u5728\u53e5\u5b50\u4e2d\u8a55\u8ad6\u9019\u4e9b\u76ee\u6a19\u7684\u4fee\u98fe\u95dc</td></tr><tr><td colspan=\"2\">\u6b32\u8a55\u8ad6\u7684\u4e8b\u4ef6\uff0c\u4f8b\u5982: \"I lovvvvvvvvve this movie.\"\u3002\u7136\u800c\u932f\u8aa4\u7684\u62fc\u5b57\u53ef\u80fd\u6703\u5f71\u97ff\u5b57\u8a5e\u4fee \u98fe\u95dc\u4fc2\u7684\u5224\u65b7\uff0c\u56e0\u6b64\u672c\u8ad6\u6587\u4f7f\u7528 Google Spell Check[25]\u5c0d\u6bcf\u5247 tweet \u9032\u884c\u62fc\u5b57\u7684\u6aa2\u67e5\uff0c \u5728\u6b64\u90e8\u4efd\u6211\u5011\u662f\u505a\u7d14\u62fc\u5b57\u932f\u8aa4\u7684\u6aa2\u67e5\uff0c\u5c0d\u65bc\u6587\u6cd5\u53ca\u5b57\u7fa9\u4f7f\u7528\u4e0d\u7576\u4e26\u7121\u505a\u6aa2\u67e5\u3002 \u4fc2\uff0c\u4e26\u6839\u64da\u60c5\u7dd2\u8a5e\u5178\u6bd4\u5c0d\uff0c\u8a08\u7b97\u51fa\u5c0d\u8a55\u8ad6\u76ee\u6a19\u7684\u60c5\u7dd2\u5206\u6578: \u984c\u7684\u8a55\u8ad6\u6216\u6558\u8ff0\uff0c\u4f8b\u5982:I \u4e2d\u96fb\u8a71\u97ff\u8d77\u800c\u611f\u5230\u4e1f\u81c9\uff0c\u6b64\u5167\u5bb9\u4e26\u6c92\u6709\u91dd\u5c0dArgo\u9019\u90e8\u96fb\u5f71\u505a\u4efb\u4f55\u7684\u8a55\u50f9\u53ca\u8ad6\u8ff0\u3002 (1) \u3001\u4fee\u98fe\u95dc\u4fc2\u8fa8\u8b58</td></tr><tr><td colspan=\"2\">\u70ba\u4e86\u80fd\u7cbe\u78ba\u8a08\u7b97\u51fa\u4f7f\u7528\u8005\u5728Twitter\u4e2d\u5c0d\u4e3b\u984c\u7684\u8a55\u50f9\uff0c\u9996\u5148\u6211\u5011\u8981\u627e\u5230tweet\u4e2d\u4f7f\u7528 \u672c\u65b9\u6cd5\u5728\u641c\u5c0b\u4fee\u98fe\u95dc\u4fc2\u6642\u90fd\u662f\u4ee5\u53e5\u5b50\u70ba\u55ae\u4f4d\uff0c\u591a\u500b\u55ae\u5b57\u653e\u5728\u4e00\u8d77\u53ef\u4ee5\u8868\u793a\u51fa\u5b8c\u6574\u7684 2\u3001 Stemming \u8005\u53ef\u80fd\u5728\u8a55\u8ad6\u7684\u4e8b\u4ef6\uff0c\u4e26\u4e14\u8981\u78ba\u8a8d\u6b64\u4e8b\u4ef6\u662f\u5426\u8ddf\u4e3b\u984c\u76f8\u95dc\u3002\u4f8b\u5982: \"I watched battleship \u610f\u601d\uff0c\u4e14\u901a\u5e38\u4ee5\u6a19\u9ede\u7b26\u865f\u5c07\u53e5\u5b50\u9593\u505a\u5340\u9694\u3002</td></tr><tr><td colspan=\"2\">\u7531\u65bc\u540d\u8a5e\u7684\u55ae\u8907\u6578 (\u5982 movie \u548c movies)\u3001\u8a5e\u6027\u7684\u8b8a\u5316 (\u5982 good \u548c goodness\uff0c\u52d5 last night, Rihanna's acting is amazing.\" \u4f8b\u5b50\u4e2d\uff0c\u5728\u8b1b\u8ff0battleship\u9019\u90e8\u96fb\u5f71\u4e2d\u6f14\u54e1\u7684\u6f14\u6280 \u4e00\u500b\u53e5\u5b50\u7684\u57fa\u672c\u7d50\u69cb\u5305\u542b\u5169\u500b\u91cd\u8981\u7684\u90e8\u5206\uff1a\u4e3b\u8a5e\u90e8\u5206(subject group)\u548c\u8ff0\u8a9e\u90e8\u5206</td></tr><tr><td colspan=\"2\">\u8a5e\u7684\u6642\u614b (\u5982 see \u548c seeing)\uff0c\u5c0e\u81f4\u8a9e\u610f\u5927\u81f4\u76f8\u540c\u7684\u8a5e\u6216\u5b57\u537b\u6709\u4e0d\u540c\u7684\u5448\u73fe\u65b9\u5f0f\uff0c\u70ba\u4e86\u8981 \u5f88\u4e0d\u932f\uff0c\u7531\u6b64\u4f8b\u53ef\u767c\u73fe\uff0c\u767c\u6587\u8005\u4e26\u4e0d\u76f4\u63a5\u8a55\u8ad6battleship\uff0c\u800c\u662f\u5c0d\u6f14\u54e1\u7684\u6f14\u6280\u505a\u597d\u7684\u8a55\u50f9\uff0c (predicate group)\uff0c\u4ea6\u5373\u4e00\u500b\u53e5\u5b50\u5fc5\u9808\u8981\u6709\u4e3b\u8a5e\u548c\u8ff0\u8a9e\u52d5\u8a5e\u3002\u800c\u610f\u898b\u8a5e\u5e38\u662f\u52d5\u8a5e\u4ee5\u53ca</td></tr><tr><td colspan=\"2\">\u7c21\u5316\u53e5\u5b50\u7684\u8907\u96dc\u7a0b\u5ea6\uff0c\u672c\u7814\u7a76\u4f7f\u7528 Porter Stemming algorithm \u4f86\u9032\u884c\u5b57\u6839\u9084\u539f\u7684\u8655\u7406\u3002 \u9019\u662f\u4e00\u7bc7\u5c0dbattleship\u6b63\u9762\u8a55\u50f9\u7684tweet\uff0c\u56e0\u70ba\u6f14\u54e1\u7684\u540d\u5b57\u4e5f\u662f\u96fb\u5f71\u7684\u5c6c\u6027\u4e4b\u4e00\u3002\u76ee\u6a19 \u5f62\u5bb9\u8a5e[3]\uff0c\u6240\u4ee5\u672c\u8ad6\u6587\u50c5\u627e\u53e5\u5b50\u4e2d\u52d5\u8a5e\u53ca\u5f62\u5bb9\u8a5e\u8207\u76ee\u6a19\u7684\u4fee\u98fe\u95dc\u4fc2[12][13]\u3002\u70ba\u4e86\u8981\u8fa8</td></tr><tr><td colspan=\"2\">\u5e0c\u671b\u5c07\u9019\u4e9b\u5f8c\u7db4\u53bb\u9664\u540c\u6642\u4e26\uf967\u5f71\u97ff\u6587\u5b57\u672c\u8eab\u7684\u610f\u7fa9\uff0c\u800c\u4e14\u5c0d\u65bc\u6aa2\u7d22\u67e5\u5168\u7387\u7684\u63d0\u5347\u4e5f\uf901\u6709 (target)\u70ba\u8207\u4e3b\u984c\u9ad8\u5ea6\u76f8\u95dc\u7684\u5b57\u8a5e\uff0c\u53ef\u80fd\u662f\u540c\u7fa9\u8a5e\u6216\u8a55\u8ad6\u4e3b\u984c\u4f7f\u7528\u7684\u5b57\u8a5e\uff0c\u672c\u5c0f\u7bc0\u5c07\u4ecb\u7d39 \u5225\u76ee\u6a19\u7684\u8a55\u8ad6\u8207\u5176\u4ed6\u6558\u8ff0\uff0c\u6211\u5011\u8a02\u5b9a\u4ee5\u4e0b\u4fee\u98fe\u898f\u5247\uff0c\u82e5\u53e5\u5b50\u4e2d\u5305\u542b\u4ee5\u4e0b\u7684\u4fee\u98fe\u95dc\u4fc2\uff0c\u5247</td></tr><tr><td colspan=\"2\">\u5e6b\u52a9\u3002 \u6211\u5011\u627etarget\u7684\u65b9\u6cd5\u3002 \u6b64\u53e5\u5b50\u53ef\u80fd\u542b\u6709\u5c0d\u76ee\u6a19\u7684\u8a55\u8ad6\u3002\u6211\u5011\u5c07\u610f\u898b\u8a5e\u7684\u610f\u898b\u50be\u5411\u53ca\u5206\u6578\u4f5c\u70ba\u6b64\u4fee\u98fe\u95dc\u4fc2\u7684\u610f\u898b</td></tr><tr><td colspan=\"2\">\u50be\u5411\u53ca\u5206\u6578: 3\u3001 \u7279\u5fb5\u904e\u6ffe (1) \u3001\u547d\u540d\u5be6\u9ad4\u8fa8\u8b58 1. VB/ VBD/ VBG/ VBN/ VBP/ VBZ (\u610f\u898b\u8a5e\uff0c\u52d5\u8a5e) + T: T \u70ba target\uff0c\u4e5f\u662f\u52d5\u8a5e\u4e4b\u5f8c\u7684</td></tr><tr><td colspan=\"2\">Tweet\u9664\u4e86\u5167\u5bb9\u672c\u6587\u4e4b\u5916\uff0c\u9084\u5305\u542b\u4ee5\u4e0b\u5e7e\u9ede\u7279\u5fb5: \uf0d8 Username: \u7d66\u4e00\u500bTweet\u7684\u56de\u8986\u6216\u7559\u8a00\u3002\u7528\u6cd5\u70ba\u5728 @ \u7b26\u865f\u5f8c\u52a0\u5c0d\u65b9\u7684 Twitter \u88dc\u8a9e\uff0c\u9867\u540d\u601d\u7fa9\uff0c\u5c31\u662f\u91dd\u5c0d\u52d5\u8a5e\uff0c\u518d\u591a\u4f5c\u63cf\u8ff0\uff0c\u88dc\u5145\u52d5\u8a5e\u4e0d\u8db3\u4e4b\u8655\uff0c\u8868\u9054\u51fa\u53e5\u5b50\u5b8c \u5728\u672c\u7814\u7a76\u4e2d\u4ee5\u96fb\u5f71\u70ba\u4f8b\uff0c\u56e0\u70ba\u6f14\u54e1\u3001\u5c0e\u6f14\u3001\u7de8\u5287\u7b49\u4eba\u7269\u90fd\u6975\u53ef\u80fd\u662f\u8a55\u8ad6\u96fb\u5f71\u7684\u7db2\u53cb \u53ef\u80fd\u8a55\u8ad6\u7684\u76ee\u6a19\uff0c\u5c08\u6709\u540d\u8a5e\u7684\u6a19\u8a18\uff0c\u53ef\u4ee5\u89e3\u6c7a\u8a5e\u5eab\u6db5\u84cb\u4e0d\u8db3\u7684\u554f\u984c\uff0c\u4e5f\u56e0\u5176\u727d\u6d89\u5230\u4eba\u3001 \u6574\u7684\u610f\u601d\uff0c\u88dc\u8db3\u65b9\u5f0f\u901a\u5e38\u662f\u4ee5\u540d\u8a5e\u6216\u4ee3\u540d\u8a5e\u4f5c\u70ba\u52d5\u8a5e\u7684\u53d7\u8a5e\u3002\u4f8b\u5982:I love battleship.\uff0c</td></tr><tr><td colspan=\"2\">ID \uff0c\u4e00\u500b\u7a7a\u683c\u6216\u5192\u865f\u5f8c\u5beb\u4e0a\u56de\u8986\u5167\u5bb9\u3002\u4f8b\u5982\uff0c\"@disc the tall man is such a good \u4e8b\u3001\u6642\u3001\u5730\u3001\u7269\u7b49\u91cd\u8981\u5167\u5bb9\uff0c\u6211\u5011\u4f7f\u7528 Stanford Named Entity Recognizer[27]\u4f86\u505a\u5c08\u6709\u540d \"love\"\u662f\u53e5\u5b50\u4e2d\u7684 VB\uff0c\"battleship\"\u662f\u96fb\u5f71\u540d\u5b57\u4e5f\u662f\u6211\u5011\u7684 target\uff0c\u5728\u6b64\u53e5\u5b50\u4e2d\u5c31\u662f</td></tr><tr><td colspan=\"2\">movie.\"\u3002 \u8a5e\u6a19\u8a18\uff0c\u4e3b\u8981\u662f\u8981\u627e\u51fa tweet \u4e2d\u53ef\u80fd\u51fa\u73fe\u8ddf\u96fb\u5f71\u6709\u95dc\u7684\u5c08\u6709\u540d\u8a5e\uff0c\u4f8b\u5982 : \u6f14\u54e1\u3001\u5c0e\u6f14\u3001</td></tr><tr><td colspan=\"2\">\uf0d8 Links (url): \u4f7f\u7528\u8005\u5e38\u6703\u5728tweet\u4e2d\u5206\u4eab\u93c8\u7d50\uff0c\u4f8b\u5982:\"That Blade Runner sequel is still \u7de8\u5287\u3001\u5176\u4ed6\u96fb\u5f71\u5c08\u696d\u8853\u8a9e\u7b49\u3002</td></tr><tr><td colspan=\"2\">happening. \u4ee5\u4e0a\u7279\u5fb5\u4e26\u4e0d\u5f71\u97ff\u4f7f\u7528\u8005\u5728tweet\u4e2d\u6b32\u8868\u9054\u7684\u6558\u8ff0\u5167\u5bb9\uff0c\u4f46\u6703\u4f7f\u5f97\u8a0a\u606f\u5167\u5bb9\u8907\u96dc\u800c\u5f71\u97ff (2) \u3001\u5171\u540c\u51fa\u73fe\u95dc\u4fc2 (Co-Occurrence)</td></tr><tr><td>\u5230\u610f\u898b\u5206\u6790\u7684\u6e96\u78ba\u7387\uff0c\u6240\u4ee5\u6211\u5011\u5c07\u9019\u4e9b\u7279\u5fb5\u4e88\u4ee5\u522a\u9664\uff0c\u53ea\u7559\u4e0b\u6558\u8ff0\u5167\u5bb9\u3002</td><td/></tr><tr><td>4\u3001 \u8a5e\u6027\u6a19\u8a3b (POS Tagging) PMI(w1,w2) =</td><td>(1)</td></tr><tr><td colspan=\"2\">specific resource)\u9032\u884c\u76ee\u6a19\u767c\u6398 (target finding)\u3002\u4ee5\u96fb\u5f71\u70ba\u4f8b\uff0c\u6211\u5011\u5f9e\u5168\u7403\u6700\u5927\u7684 \u56e0\u70ba\u7814\u7a76\u4e2d\u9808\u627e\u51fa\u8a5e\u8207\u8a5e\u4e2d\u7684\u4fee\u98fe\u53ca\u5c0d\u7b49\u95dc\u4fc2\uff0c\u4efb\u4f55\u8a9e\u8a00\u8655\uf9e4\u7684\u7cfb\u7d71\ufa26\u5fc5\u9808\u5148\u5206\u8fa8 \u5982\u516c\u5f0f 1 \u6240\u793a\uff0c P(w1)\u548c P(w2)\u53ef\u4ee5\u900f\u904e\u8a08\u7b97 w1 \u548c w2 \u500b\u5225\u51fa\u73fe\u7684\u6b21\uf969\u4f5c\u70ba\u6a5f\uf961\u4f30</td></tr><tr><td colspan=\"2\">\u96fb\u5f71\u67e5\u8a62\u8cc7\u6599\u5eab IMDB (The Internet Movie Database)\u4e2d\u6536\u96c6\u96fb\u5f71\u7684\u76f8\u95dc\u4f5c\u8005\u3001\u5c0e\u6f14\u3001\u6f14 \u6587\u672c\u4e2d\u7684\u8a5e\u624d\u80fd\u9032\ufa08\u9032\u4e00\u6b65\u7684\u8655\uf9e4\uff0c\u6211\u5011\u4f7f\u7528 Stanford POS Tagger[26]\u9032\u884c\u8a5e\u6027\u6a19\u8a3b\u3002 \u8a08\u503c\uff1b\u800c P(w1,w2)\u4ee3\u8868 w1 \u548c w2 \uf978\u500b\u5b57\u5171\u540c\u51fa\u73fe (co-occurrence)\u7684\u6a5f\uf961\uff0c\u53ef\u4ee5\u900f\u904e\u8a08</td></tr><tr><td colspan=\"2\">\u54e1\u53ca\u985e\u578b\u7b49\u8cc7\u8a0a\u3002\u63a5\u8457\u900f\u904e\u610f\u898b\u5206\u6790\u6a21\u7d44\u5206\u6790\u6bcf\u5247 tweet \u5167\u5bb9\u662f\u5426\u542b\u6709\u5c0d\u8a72\u4e3b\u984c\u76f8\u95dc\u76ee \u7b97\uf978\u500b\u5b57\u5728\u6587\u7ae0\u4e2d\u5171\u540c\u51fa\u73fe\u7684\u6b21\uf969\u4f5c\u70ba\u6a5f\uf961\u4f30\u8a08\u503c\u3002 (\u4e09) \u3001\u610f\u898b\u5206\u6790 \u6a19\u7684\u8a55\u8ad6\uff0c\u9032\u800c\u8a08\u7b97\u610f\u898b\u5206\u6578\u4f86\u5224\u65b7\u8a55\u50f9\u7684\u6b63\u8ca0\u9762\u3002 expansion)\uff0c\u610f\u898b\u8a5e\u8207\u8a55\u8ad6\u76ee\u6a19\u4fee\u98fe\u95dc\u4fc2 (opinion words modification relation)\uff0c\u6700\u5f8c\u8a08\u7b97 (\u4e8c) \u3001\u524d\u8655\u7406 \u672c\u7ae0\u7bc0\u8aaa\u660e\u610f\u898b\u5206\u6790\u7684\u65b9\u6cd5\uff0c\u4e3b\u8981\u5206\u70ba\u4e09\u500b\u6b65\u9a5f: \u767c\u6398\u76f8\u95dc\u7684\u8a55\u8ad6\u76ee\u6a19 (target (3) \u3001\u540c\u7fa9\u8a5e</td></tr><tr><td colspan=\"2\">\u70ba\u4e86\u66f4\u5bb9\u6613\u627e\u5230\u8a0a\u606f\u4e2d\u7684\u610f\u898b\u8a5e (opinion word)\u53ca\u8a55\u8ad6\u4e3b\u984c (topic)\uff0c\u6211\u5011\u5c07\u8207\u4e3b\u984c \u53e5\u5b50\u7684\u610f\u898b\u5206\u6578 (opinion Score estimation)\u4f86\u5224\u65b7\u53e5\u5b50\u7684\u610f\u898b\u50be\u5411 (opinion orientation</td></tr><tr><td>\u76f8\u95dc\u7684 tweet \u505a\u524d\u7f6e\u8655\u7406\u4f86\u9054\u5230\u7c21\u5316\u53e5\u5b50\u7684\u76ee\u7684\u3002 identification)\u3002</td><td/></tr></table>", | |
| "type_str": "table", | |
| "num": null | |
| }, | |
| "TABREF3": { | |
| "html": null, | |
| "text": "\u4e2d\u7684\u610f\u898b\u8a5e\uff0cT \u70ba\u7d93\u7531 target finding \u6240\u627e\u5230 target \u7684\u96c6\u5408\uff0cd(opj,t i ) \u662f\u5728\u53e5\u5b50 s \u4e2d\u610f\u898b\u8a5e opj \u53ca ti \u7684\u8ddd\u96e2\uff0cso \u662f\u4fee\u98fe\u5b57 opj \u7684\u60c5\u7dd2\u9762\u5411\u5206\u6578\uff0c\u7531 SentiWordNet \u5f97\u77e5\u3002\u516c\u5f0f\u7684 multiplicative inverse \u662f\u70ba\u4e86\u5224\u65b7\u4fee\u98fe\u5b57\u5728 \u4fee\u98fe target \u7684\u53ef\u80fd\u6027\uff0c\u82e5\u8ddd\u96e2 \u8d8a\u9060\u5247\u8a08\u7b97\u51fa\u7684\u60c5\u7dd2\u5206\u6578\u8d8a\u4f4e\u3002\u6574\u7bc7 tweet \u7684\u5206\u6578\u5373\u70ba\u6240\u6709\u53e5\u5b50\u60c5\u7dd2\u5206\u6578\u7684\u7e3d\u548c\u3002\u6700\u5f8c \u4f9d\u64da\u5206\u6578\u5c07 tweet \u5206\u6210\u4e09\u985e: NER \u5728\u64f7\u53d6\u4eba\u540d\u53ca\u5c08\u6709\u540d\u8a5e\u6642\u53ef\u80fd\u6703\u627e\u5230\u7121\u95dc\u7684 target \uff0c\u56e0\u6b64\u5728 precision \u53ea\u6709\u7a0d\u5fae\u9032\u6b65\u3002 2. \u5171\u540c\u51fa\u73fe\u95dc\u4fc2 \u5982\u5716\u4e8c\u3001\u5716\u4e09\u6240\u793a\uff0c\u6211\u5011\u5206\u5225\u89c0\u5bdf\u4e3b\u5ba2\u89c0\u53ca\u6b63\u9762\u60c5\u7dd2\u8a55\u8ad6\u7684\u5206\u985e\u6548\u679c\uff0c\u7576 k \u503c\u7531 0 \u5230 15 \u6642\uff0c\u56e0\u70ba\u589e\u52a0 target \u7684\u6578\u91cf\uff0crecall \u662f\u6301\u7e8c\u4e14\u660e\u986f\u7684\u4e0a\u5347\uff0cprecision \u53ef\u80fd\u56e0\u70ba\u7576\u8003 \u616e\u7684 opinion target \u589e\u52a0\u4f7f\u5f97\u4fee\u98fe\u95dc\u4fc2\u8b8a\u5f97\u8907\u96dc\u800c\u5448\u73fe\u6301\u7e8c\u4e0b\u964d\u7684\u8da8\u52e2\uff0c\u4f46 accuracy \u7684 \u503c\u4e5f\u56e0\u70ba k \u503c\u7684\u589e\u52a0\u800c\u6709\u660e\u986f\u4e0a\u5347\u3002\u7576 k \u503c\u7531 16 \u5230 23 \u6642\uff0ctarget \u8207\u96fb\u5f71\u7684\u76f8\u95dc\u5ea6\u4e0b\u964d\uff0c precision \u4ecd\u7136\u662f\u5448\u73fe\u6301\u7e8c\u4e0b\u964d\u7684\u8da8\u52e2\uff0c\u800c recall \u4e26\u6c92\u6709\u518d\u4e0a\u5347\uff0c\u6240\u4ee5 F1 score \u4ecd\u7136\u5728\u6301 \u7e8c\u4e0b\u964d\u3002accuracy \u503c\u4e5f\u56e0\u70ba\u9810\u6e2c\u5931\u6557\u9010\u6f38\u4e0b\u964d\u3002 \u5716\u4e8c\u3001\u4e0d\u540c k \u503c PMI \u5c0d\u4e3b\u5ba2\u89c0\u8a55\u8ad6\u5206\u985e\u7684 precision\u3001recall \u53ca accuracy \u5f71\u97ff \u5716\u4e09\u3001\u4e0d\u540c k \u503c PMI \u5c0d\u6b63\u9762\u60c5\u7dd2\u5206\u985e\u7684 precision\u3001recall \u53ca accuracy \u5f71\u97ff \u6839\u64da\u8868\u56db\u3001\u8868\u4e94\u6211\u5011\u89c0\u5bdf\u5230 precision \u662f\u4e0b\u964d\u7684\uff0c\u7576\u53c3\u8003\u8d8a\u591a\u7684 target\uff0c\u5c31\u6703\u589e\u52a0\u53e5 \u6240\u63d0\u65b9\u6cd5\u8207 SVM \u53ca Na\u00efve Bayes Classifier \u4e4b\u6bd4\u8f03\u8207\u8a0e\u8ad6 \u6211\u5011\u5229\u7528 n-gram \u5c07\u8a0a\u606f\u5167\u5bb9\u4f5c\u5207\u5272\uff0c\u6240\u627e\u5230\u7684\u5206\u5272\u5b57\u7576\u4f5c\u4e00\u7d44\u7368\u7acb\u7684\u8a5e\u5f59\uff0c\u56e0\u70ba \u8a0a\u606f\u5167\u5bb9\u8f03\u77ed\uff0c\u6211\u5011\u4f7f\u7528 unigram \u548c bigram\u3002\u8a0a\u606f\u900f\u904e n-gram \u65b7\u8a5e\u6f14\u7b97\u6cd5\u5f97\u51fa\u7684 n-gram \u7279\u5fb5\u503c\u5957\u5165\u652f\u6301\u5411\u91cf\u6a5f(Support Vector Machine, SVM)\u53ca\u8c9d\u6c0f\u5206\u985e\u5668(Na\u00efve Twitter \u5728\u767c\u6587\u7684\uf969\u5b57\u4e0a\u6709\u9650\u5236\uff0c\u4e3b\u8981\u662f\u4ee5\u5e73\u5e38\u53e3\u8a9e\u3001\u7c21\u77ed\u7684\u65b9\u5f0f\u5728 Twitter \u4e0a\u767c \u4f48\uff0c\u56e0\u6b64\u5e38\u6709\u4fda\u8a9e\u7684\u90e8\u5206\uff0c\u6240\u4ee5\u5728\u5b57\u4e32\u8655\u7406\u65b9\u9762\u6703\u6709\u56f0\u96e3\u3002\u4f8b\u5982: \"shh identity thief, is it good movie I gonna bring my Lil g Cuzco to DE movies, Rollin out Tass can't wait till class finish lol.\"\uff0c\u51fa\u73fe\u5f88\u591a\u7c21\u5beb\u53ca\u53e3\u8a9e\u5316\u7684\u5b57\uff0c\u5728\u524d\u8655\u7406\u6642\u7684\u5b57\u4e32\u8655\u7406\u9020\u6210\u56f0\u96e3\uff0c\u4ee5\u81f3\u65bc\u5f71 \u97ff\u8a5e\u6027\u6a19\u8a3b\u3001\u8a55\u8ad6\u76ee\u6a19\u767c\u6398\u4ee5\u53ca\u4fee\u98fe\u95dc\u4fc2\uff0c\u6240\u4ee5\u6703\u964d\u4f4e\u672c\u65b9\u6cd5\u7684\u6e96\u78ba\u7387\u3002\u76ee\u524d\u4e5f\u6c92\u6709\u8f03 \u6b63\u5f0f\u7684 tweet \u96fb\u5f71\u8a55\u8ad6\u8cc7\u6599\u96c6\uff0c\u672c\u5be6\u9a57\u662f\u81ea\u5df1\u6536\u96c6\u8cc7\u6599\u96c6\u4e26\u4e14\u5229\u7528\u4eba\u5de5\u6a19\u8a3b\u65b9\u6cd5\u4f86\u505a\u8cc7 \u6599\u96c6\u7684\u60c5\u7dd2\u6a19\u8a3b\uff0c\u5728\u6578\u91cf\u4e0a\u8f03\u70ba\u4e0d\u8db3\u3002\u4ee5\u4e0a\u7684\u96e3\u984c\u90fd\u662f\u672a\u4f86\u5f85\u514b\u670d\u7684\u8b70\u984c\u3002 \u53c3\u8003\u6587\u737b", | |
| "content": "<table><tr><td>__________________________________ (\u4e8c) \u3001\u5be6\u9a57\u7d50\u679c\u8207\u8a0e\u8ad6</td></tr><tr><td>(3) \u3001\u5426\u5b9a\u8a5e \u6839\u64da Tottie[16]\uff0c\u82f1\u6587\u7684\u5426\u5b9a\u6a19\u8a18 (negative marker)\u4e3b\u8981\u5206\u70ba\u4e09\u5927\uf9d0\uff1a (1) not \u5426\u5b9a(not-negation) (2) no \u5426\u5b9a(no-negation) (3) \u8a5e\u7db4\u5426\u5b9a(affixal negation) \u5426\u5b9a\u6a19\u8a18\u7684\u7bc4\u4f8b\u5982\u8868\u4e00\u6240\u793a\uff1a \u8868\u4e00\u3001\u5426\u5b9a\u6a19\u8a18 ________________________________ not-negation no-negation affixal negation not No (im)perfect nor (ir)respective none (in)dependent never (un)able neither (non)functional \u7531 Tottie \u7684\u5b9a\u7fa9\u4e2d\u53ef\u4ee5\u767c\u73fe\uff0cnot \u5426\u5b9a\u548c no \u5426\u5b9a\u57fa\u672c\u4e0a\u5c6c\u65bc\u8a9e\u6cd5\u7684\u7bc4\u7587\uff0c\u800c\u8a5e\u7db4\u5426\u5b9a (affixal negation)\u5247\u662f\u5728\u8a5e\u5f59\u7684\u7bc4\u7587\u3002\u5728\u8a5e\u7db4\u5426\u5b9a\u7684\u90e8\u5206\u5728 SentiWordNet \u80fd\u4f5c\u9069\u7576\u7684 \u8fa8\u5225\uff0c\u4f8b\u5982 : perfect positive \uff0c imperfect negative\u3002 not-negation \u8207 no-negation \u7684\u8655\u7406[17]\uff0c\u5148\u4f9d\u7167\u5148\u524d\u4ecb\u7d39\u7684\u65b9\u6cd5\u627e\u5230\u4fee\u98fe\u95dc\u4fc2\u53ca\u6b63\u8ca0\u9762 \u60c5\u7dd2\uff0c\u82e5\u662f\u53e5\u5b50\u4e2d\u542b\u6709 not-negation \u8207 no-negation \u7684\u5b57\uff0c\u5247\u6703\u53cd\u8f49\u6b63\u8ca0\u9762\u7d50\u679c\uff0c\u4f8b\u5982: I don't like this movie, the plot is so boring. \u4f8b\u5b50\u4e2d\uff0c\u4f9d\u7167\u4e4b\u524d\u4ecb\u7d39\u7684\u898f\u5247\u5728\" I don't like this movie,\"\u53e5\u5b50\u4e2d\u7684\"like this movie\"\u627e\u5230 VB + Target \u7684\u4fee\u98fe\u95dc\u4fc2\uff0c\u5c6c\u65bc\u6b63\u9762\u8a55\u50f9\u7684\u53e5 \u5b50\uff0c\u4f46\u5728\u53e5\u4e2d\u627e\u5230\"n't\"\u7684\u5426\u5b9a\u8a5e\uff0c\u6240\u4ee5\u539f\u5c6c\u6b63\u9762\u8a55\u50f9\u7684\u53e5\u5b50\u5728\u6700\u5f8c\u6703\u53cd\u8f49\u6210\u8ca0\u9762\u8a55\u50f9\u3002 3. \u610f\u898b\u8a55\u5206 Tweets \u5728\u7d93\u904e\u524d\u7ae0\u7bc0\u7684\u4fee\u98fe\u95dc\u4fc2\u7279\u5fb5\u7684\u641c\u5c0b\uff0c\u9019\u4e9b\u4fee\u98fe\u95dc\u4fc2\u6703\u56e0\u70ba\u5b57\u8207\u5b57\u4e4b\u9593\u7684 \u8ddd\u96e2\u800c\u5f71\u97ff\u5230\u60c5\u7dd2\u5206\u6578\uff0c\u4f8b\u5982: In the first movie Tony Curtis's acting is amazing.\uff0c\u4f8b\u5b50 \u4e2d\uff0c\u627e\u5230\"T + VBZ + JJ\"\u7684\u4fee\u98fe\u95dc\u4fc2\"acting + is + amazing\"\uff0c\u6211\u5011\u6703\u8a08\u7b97\u610f\u898b\u8a5e\u8207 target \u4e4b\u9593\u7684\u8ddd\u96e2\u4f86\u8abf\u6574\u4fee\u98fe\u7684\u6b0a\u91cd\u3002\u4e00\u5247 tweet \u4e2d\u53ef\u80fd\u5b58\u5728\u8a31\u591a\u4fee\u98fe\u95dc\u4fc2\u7684\u7279\u5fb5\uff0c\u6240\u4ee5\u9700\u8981 \u7d93\u904e\u6b63\u8ca0\u9762\u60c5\u7dd2\u5206\u6578\u7684\u52a0\u7e3d\u4f86\u5224\u65b7\u6b64 tweet \u662f\u5c6c\u65bc\u6b63\u9762\u60c5\u7dd2\u6216\u8ca0\u9762\u60c5\u7dd2\uff0c\u4ea6\u6216\u662f\u5ba2\u89c0\u8ad6 \u8ff0\u7684 tweet\u3002\u91dd\u5c0d\u67d0\u53e5\u5b50 s\uff0c\u5176\u60c5\u7dd2\u5206\u6578\u7684\u8a08\u7b97\u5982\u4e0b: score(s) = \uff0c (2) \u516c\u5f0f\u4e2d\uff0copj\u662f\u53e5\u5b50 s \uf0d8 \u6b63\u9762(positive): score > 0\u3002 \uf0d8 \u8ca0\u9762(negative): score < 0\u3002 \uf0d8 \u5ba2\u89c0(objectivity): score = 0\u3002 \u56db\u3001\u5be6\u9a57\u8207\u8a0e\u8ad6 (\u4e00) \u3001\u6e2c\u8a66\u8cc7\u6599\u6536\u96c6 \u96a8\u6a5f\u6311\u9078\u5728 2013 \u5e74 2 \u6708\u81f3 3 \u6708\u4e0a\u6620\u7684\u4e94\u5341\u90e8\u96fb\u5f71\u6536\u96c6\u5176\u76f8\u95dc\u8a55\u8ad6 tweet\u3002\u56e0\u70ba\u4e0d\u60f3 \u4f7f\u8cc7\u6599\u904e\u65bc\u96c6\u4e2d\u5728\u67d0\u4e00\u5929\uff0c\u6240\u4ee5\u6bcf\u9694 5 \u5929\u6536\u96c6\u4e00\u6b21\u3002\u6536\u96c6\u65e5\u671f\u5206\u5225\u70ba 2013/3/20\u3001 2013/3/25\u30012013/3/30\u30012013/4/5 \u53ca 2013/4/10\uff0c\u6bcf\u65e5\u7684\u6536\u96c6\u91cf\u70ba 200 \u5247 tweets\uff0c\u6e2c\u8a66\u8cc7\u6599 \u7e3d\u5171 1000 \u5247 tweets\uff0c\u63a5\u8457\u4f7f\u7528\u4eba\u5de5\u6a19\u8a3b\u6bcf\u5247 tweet \u7684\u60c5\u7dd2\u9762\u5411\u4f5c\u70ba\u5be6\u9a57\u7684\u6a19\u6e96\u7b54\u6848\uff0c\u7531 5 \u4eba\u9032\u884c\u60c5\u7dd2\u6a19\u8a3b\uff0c\u6a19\u8a3b\u6709\u4e09\u985e:\u6b63\u9762\u3001\u8ca0\u9762\u3001\u5ba2\u89c0\u3002\u82e5\u662f\u6b63\u9762\u70ba+1\uff0c\u8ca0\u9762\u70ba-1\uff0c\u5ba2\u89c0\u70ba \u8868\u56db\u3001\u52a0\u5165 PMI (k=15)\u524d\u5f8c\u7684\u4e3b\u5ba2\u89c0\u8a55\u8ad6\u5206\u985e\u6548\u679c\u6bd4\u8f03 \u6211\u5011\u5728\u9032\u884c\u7814\u7a76\u65b9\u6cd5\u4e2d\u7684 target expansion \u7684\u5be6\u9a57\u8207\u8a0e\u8ad6\uff0c\u6700\u5f8c\u6bd4\u8f03\u672c\u8ad6\u6587\u7684\u65b9\u6cd5\u8207 SVM \u53ca Naive Bayes \u5206\u985e\u65b9\u6cd5\u7684\u6548\u679c\u3002\u5206\u5225\u6703\u8a08\u7b97\u51fa\u6b63\u8ca0\u9762\u53ca\u4e3b\u89c0(subjectivity)\u8a55\u8ad6\u7684 \u7cbe\u78ba\uf961(precision)\u3001\u67e5\u5168\u7387(recall)\u3001F1 \u53ca\u6e96\u78ba\uf961(accuracy)\u7b49\u6578\u503c\u4f86\u8861\uf97e\u65b9\u6cd5\u6548\u679c\u3002 \u5176\u4e2d baseline \u70ba\u672a\u7d93\u904e\u8a55\u8ad6\u76ee\u6a19\u767c\u6398\uff0c\u6240\u4ee5 target \u53ea\u5305\u542b\u96fb\u5f71\u540d\u5b57\uff0c\u85c9\u6b64\u4f86\u6bd4\u8f03\u8a55 \u8ad6\u76ee\u6a19\u767c\u6398\u65b9\u6cd5\u7684\u6548\u679c\u3002 1. \u547d\u540d\u5be6\u9ad4\u8fa8\u8b58 Baseline \u56e0\u70ba target \u904e\u65bc\u7a00\u5c11\uff0c\u4f7f\u5f97 recall \u90fd\u904e\u65bc\u504f\u4f4e\uff0c\u7121\u6cd5\u6709\u6548\u7684\u8fa8\u8b58\u5927\u90e8\u5206\u6709 \u95dc\u96fb\u5f71\u7684\u8a55\u8ad6\u3002\u96fb\u5f71\u8a55\u8ad6\u7684 target \u53ef\u80fd\u4e5f\u5305\u542b\u4eba\u540d\u53ca\u5c08\u6709\u540d\u8a5e\uff0c\u6211\u5011\u4f7f\u7528 Stanford Named Entity Recognition \u5de5\u5177\u64f4\u589e\u8a55\u8ad6\u76ee\u6a19\u7684\u6578\u91cf\u3002\u52a0\u5165 NER \u524d\u5f8c\u7684\u5206\u985e\u6548\u679c\u5982\u8868\u4e8c\u3001\u8868\u4e09 \u6240\u793a\u3002 \u8868\u4e8c\u3001\u52a0\u5165 Named Entity Recognition \u524d\u5f8c\u7684\u4e3b\u5ba2\u89c0\u8a55\u8ad6\u5206\u985e\u6548\u679c\u6bd4\u8f03 \u4e3b\u89c0\u5206\u985e Baseline with Named Entity Recognition Improvement (%) Recall 0.38173 0.40315 5.6% Precision 0.89247 0.90671 1.6% F score 0.53474 0.55814 4.4% Accuracy 0.57438 0.59594 3.8% \u96d6\u7136\u4f7f\u7528 Named Entity Recognition \u6703\u589e\u52a0\u96fb\u5f71\u76f8\u95dc\u7684\u5c0e\u6f14\u3001\u6f14\u54e1\u7b49\u7684\u540d\u5b57\u53ca\u5c08\u6709\u540d\u8a5e\uff0c \u4f46\u4e5f\u6703\u4f7f\u8207\u96fb\u5f71\u7121\u95dc\u7684\u540d\u5b57\u53ca\u5c08\u6709\u540d\u8a5e\u4e5f\u6703\u7d0d\u5165 target\uff0c\u4f46\u56e0\u70ba\u6211\u5011\u6839\u64da\u96fb\u5f71\u540d\u7a31\u53bb\u6536 \u96c6 tweet\uff0c\u6240\u4ee5\u5927\u90e8\u4efd\u7684 tweet \u5167\u5bb9\u90fd\u662f\u5728\u8a55\u8ad6\u96fb\u5f71\uff0c\u5982\u8868\u4e8c\u6240\u793a\uff0c\u5728\u5224\u65b7\u4e3b\u5ba2\u89c0\u8a55\u8ad6\u6703 \u56e0\u70ba target \u7684\u589e\u52a0\u5728 precision\u3001recall \u53ca accuracy \u90fd\u6709\u63d0\u5347\u3002 \u8868\u4e09\u3001\u52a0\u5165 Named Entity Recognition \u524d\u5f8c\u7684\u6b63\u8ca0\u9762\u8a55\u8ad6\u5206\u985e\u6548\u679c\u6bd4\u8f03 \u6b63\u8ca0\u9762\u8a55\u8ad6 Baseline with Named Entity Recognition Improvement (%) Positive Recall 0.402 0.41673 3.7% Negative Recall 0.39483 0.41837 5.9% Positive Precision 0.91224 0.92194 1.1% Negative Precision 0.88563 0.90173 1.8% Positive F score 0.55807 0.57400 2.9% Negative F score 0.54616 0.57156 4.7% Accuracy 0.58813 0.60956 3.6% \u5982\u8868\u4e09\u6240\u793a\uff0c\u96a8\u8457\u4e3b\u5ba2\u89c0\u8a55\u8ad6\u5206\u985e\u7684\u6548\u679c\u63d0\u5347\uff0c\u9032\u4e00\u6b65\u5c07\u4e3b\u89c0\u8a55\u8ad6\u5206\u8fa8\u6b63\u9762(positive) \u53ca\u8ca0\u9762(negative)\u7684\u6548\u80fd\u4e5f\u80fd\u6709\u6240\u63d0\u5347\u3002 \u5b50 \u5206 \u6790 \u7684 \u8907 \u96dc \u5ea6 \u3002 \u8f03 \u7c21 \u55ae \u7684 \u53e5 \u5b50 \u80fd \u5920 \u5224 \u65b7 \u6b63 \u78ba \u4f8b \u5982 :\"Loving the music in total recall :-)\"\uff0c\u4f8b\u53e5\u662f\u5728\u8a55\u8ad6\"Total Recall\"\u9019\u90e8\u96fb\u5f71\u7684\u97f3\u6a02\u3002\u4f8b\u5b50\u4e2d\u627e\u5230\u610f\u898b\u8a5e\"loving\"\u4fee \u98fe target \"music\"\u3002\u8907\u96dc\u7684\u53e5\u5b50\uff0c\u4f8b\u5982:\" Watching Warm Bodies! :D right after i listen to my 5SOS playlist.... I have a problem... Im addicted to 5SOS music...\"\uff0c\u4f8b\u53e5\u4e2d\u51fa\u73fe \"Warm Bodies\"\u53ca\"music\"\u9019\u5169\u500b target\uff0c\u627e\u5230\u5169\u7d44\u4fee\u98fe\u95dc\u4fc2: \u4e00\u70ba Watching \u4fee\u98fe Warm Bodies\uff0c \u7cfb\u7d71\u5224\u65b7\u70ba objective\uff0c\u53e6\u4e00\u70ba addicted \u4fee\u98fe music\uff0c\u7cfb\u7d71\u5224\u65b7\u70ba positive\u3002\u6700\u5f8c\u7d93\u904e\u5206\u6578 \u7684\u52a0\u7e3d\u5224\u65b7\u6b64 tweet \u70ba positive\u3002\u7136\u800c\u6b64\u8a55\u8ad6\u4e26\u4e0d\u662f\u91dd\u5c0d\u96fb\u5f71\u7684\u97f3\u6a02\uff0c\u6240\u4ee5\u6b64 tweet \u61c9 \u4e3b\u89c0\u5206\u985e Baseline k = 15 Improvement (%) Recall 0.38173 0.68176 78.6% Precision 0.89247 0.85247 -4.4% F score 0.53474 0.75761 41.7% Accuracy 0.57438 0.79341 38.1% \u8868\u4e94\u3001\u52a0\u5165 PMI (k=15)\u524d\u5f8c\u7684\u6b63\u8ca0\u9762\u8a55\u8ad6\u5206\u985e\u6548\u679c\u6bd4\u8f03 \u6b63\u8ca0\u9762\u8a55\u8ad6 Baseline k = 15 Improvement (%) Positive Recall 0.402 0.70449 75.2% Negative Recall 0.39483 0.67968 72.1% Positive Precision 0.91224 0.87124 -4.5% Negative Precision 0.88563 0.84963 -4.1% Positive F score 0.55807 0.77904 39.6% Negative F score 0.54616 0.75521 38.3% Accuracy 0.58813 0.82732 40.7% 3. \u540c\u7fa9\u8a5e \u5982\u8868\u516d\u6240\u793a\uff0c\u7d93\u7531\u540c\u7fa9\u8a5e\u4f86\u64f4\u589e target \u7684\u6578\u91cf\uff0c\u5728\u4e3b\u5ba2\u89c0\u8a55\u8ad6\u5206\u985e\u7684\u5be6\u9a57\u6578\u503c\u90fd\u6709 \u6240\u63d0\u5347\uff0c\u539f\u56e0\u662f\u80fd\u514b\u670d\u4e0d\u540c\u4f7f\u7528\u8005\u8a55\u8ad6\u540c\u4e00\u4e8b\u7269\u537b\u6709\u5f88\u591a\u55ae\u5b57\u53ef\u4ee5\u8868\u9054\u7684\u554f\u984c\uff0c\u56e0\u6b64\u5728 tweet \u4e2d\u80fd\u627e\u5230\u66f4\u591a\u8207\u96fb\u5f71\u6709\u95dc\u7684\u8a55\u8ad6\u53ca\u4fee\u98fe\u95dc\u4fc2\u3002 \u8868\u516d\u3001\u52a0\u5165\u540c\u7fa9\u8a5e\u524d\u5f8c\u4e4b\u4e3b\u5ba2\u89c0\u8a55\u8ad6\u5206\u985e\u6548\u679c\u6bd4\u8f03 \u4e3b\u89c0\u5206\u985e PMI (k = 15) PMI (k = 15) + synonyms Improvement (%) F score 0.75761 0.77374 2.1 % Accuracy 0.79341 0.80971 2.0 % \u5982\u8868\u4e03\u6240\u793a\uff0c\u56e0\u70ba\u5224\u65b7\u4e3b\u89c0\u8a55\u8ad6\u7684\u6548\u679c\u589e\u52a0\uff0c\u5728\u6b63\u8ca0\u9762\u8a55\u8ad6\u5206\u985e\u6548\u80fd\u4e5f\u6709\u6240\u63d0\u5347\u3002 \u9019\u8aaa\u660e\u540c\u4e00\u4ef6\u4e8b\u7269\u6216\u4e8b\u4ef6\uff0c\u4e0d\u540c\u4f7f\u7528\u8005\u6703\u4f7f\u7528\u4e0d\u540c\u5b57\u773c\u4f86\u8a55\u8ad6\u6216\u6558\u8ff0\uff0c\u6240\u4ee5\u7d93\u7531\u627e\u540c\u7fa9 \u5b57\u80fd\u5f4c\u88dc\u9019\u65b9\u9762\u7684\u4e0d\u8db3\u3002 \u8868\u4e03\u3001\u540c\u7fa9\u8a5e\u524d\u5f8c\u4e4b\u6b63\u8ca0\u9762\u8a55\u8ad6\u5206\u985e\u6548\u679c\u6bd4\u8f03 \u6b63\u8ca0\u9762\u8a55\u8ad6 PMI (k = 15) PM I(k = 15) + synonyms Improvement (%) Positive F score 0.77904 0.79307 1.8 % Negative F score 0.75521 0.76201 0.9 % Accuracy 0.82732 0.83929 1.4 % \u5668\u4f7f\u7528\u5b57\u7684\u7279\u5fb5\u9032\u884c\u77ed\u8a0a\u606f\u7684\u60c5\u7dd2\u5206\u985e\u7684\u6548\u679c\u3002 \u900f\u904e\u524d\u9762\u5be6\u9a57\u7684\u89c0\u5bdf\uff0c\u767c\u73fe\u5728 target expansion \u4f7f\u7528 Named Entity Recognizer\u3001PMI (k = 15)\u53ca Synonyms \u53ef\u63d0\u5347\u5206\u985e\u7684\u6548\u679c\uff0c\u56e0\u6b64\uff0c\u9019\u88e1\u6211\u5011\u5728 target expansion \u4f7f\u7528\u4e0a\u8ff0\u65b9\u6cd5 \u4e26\u900f\u904e\u4fee\u98fe\u95dc\u4fc2\u7684\u5206\u6578\u8a08\u7b97\u9032\u884c\u5be6\u9a57\uff0c\u4e26\u8207 SVM \u53ca Naive Bayes \u5206\u985e\u7684\u6548\u679c\u6bd4\u8f03\u3002 \u5f9e\u8868\u516b\u5f97\u77e5\uff0c\u6211\u5011\u7684\u5206\u6cd5\u7121\u8ad6\u5728 positive\u3001negative \u53ca subjectivity \u7684\u7cbe\u78ba\uf961\u660e\u986f\u512a \u65bc SVM \u53ca Naive Bayes\u3002 \u8868\u516b\u3001\u6240\u63d0\u65b9\u6cd5\u8207 LibSVM \u53ca Naive Bayes Classifier precision \u7684\u6bd4\u8f03 \u6240\u63d0\u65b9\u6cd5 LibSVM Naive Bayes Positive Precision 0.88893 0.72810 0.68017 Negative Precision 0.85392 0.69724 0.63694 Subjectivity Precision 0.87269 0.69378 0.63954 \u56e0\u70ba\u672c\u7814\u7a76\u662f\u91dd\u5c0d\u8a55\u8ad6\u76ee\u6a19\u8207\u610f\u898b\u8a5e\u4e4b\u9593\u7684\u4fee\u98fe\u95dc\u4fc2\uff0c\u6240\u4ee5\u53ea\u6709\u5728\u53e5\u5b50\u4e2d\u627e\u7684\u7279\u5fb5 \u624d\u6703\u5224\u65b7\u60c5\u7dd2\u9762\u5411\uff0c\u53ef\u80fd\u7684\u76ee\u6a19\u8f03\u4fb7\u9650\uff0c\u7121\u6cd5\u627e\u51fa\u6240\u6709\u8a55\u8ad6\u8005\u53ef\u80fd\u8a55\u8ad6\u7684\u76ee\u6a19\uff0c\u6240\u4ee5\u5728 recall \u7684\u6548\u679c\u6703\u4f4e\u65bc SVM \u53ca Naive Bayes\uff0c\u5982\u8868\u4e5d\u6240\u793a\u3002 \u8868\u4e5d\u3001\u6240\u63d0\u65b9\u6cd5\u8207 LibSVM \u53ca Naive Bayes Classifier recall \u7684\u6bd4\u8f03 \u6240\u63d0\u65b9\u6cd5 LibSVM Naive Bayes Positive Recall 0.72718 0.88531 0.92423 Negative Recall 0.69796 0.87938 0.77431 Subjectivity Recall 0.71284 0.88157 0.90767 \u5f9e\u8868\u5341\u5f97\u77e5\uff0c\u672c\u7814\u7a76\u65b9\u6cd5\u7684 accuracy \u512a\u65bc SVM\uff0c\u662f\u56e0\u70ba\u6709\u8f03\u597d\u7684\u7cbe\u78ba\uf961\u3002\u6839\u64da\u5be6 \u9a57\u4e2d\u7684 precision \u53ca accuracy\uff0c\u672c\u7814\u7a76\u80fd\u5920\u91dd\u5c0d\u77ed\u8a0a\u606f\u4e2d\u7684\u67d0\u7279\u5b9a\u4e3b\u984c\u505a\u6709\u6548\u4e3b\u3001\u5ba2\u89c0\u8a55 \u8ad6\u7684\u5206\u985e\uff0c\u4e26\u4e14\u80fd\u9032\u4e00\u6b65\u5c07\u4e3b\u89c0\u8a55\u8ad6\u7cbe\u78ba\u5730\u5206\u985e\u51fa\u6b63\u3001\u8ca0\u9762\u7684\u60c5\u7dd2\u50be\u5411\u3002 \u8868\u5341\u3001\u6240\u63d0\u65b9\u6cd5\u8207 LibSVM \u53ca Naive Bayes Classifier accuracy \u7684\u6bd4\u8f03 \u6240\u63d0\u65b9\u6cd5 LibSVM Naive Bayes \u4e3b\u89c0\u5206\u985e Accuracy 0.82271 0.81673 0.78923 \u6b63\u8ca0\u9762\u8a55\u8ad6 Accuracy 0.84439 0.83439 0.81195 \u4e94\u3001\u7d50\u8ad6 \u672c\u8ad6\u6587\u63d0\u51fa\u4e00\u500b\u57fa\u65bc\u8a55\u8ad6\u76ee\u6a19\u767c\u6398\u53ca\u610f\u898b\u8a5e\u4fee\u98fe\u95dc\u4fc2\u4e4b\u5fae\u7db2\u8a8c\u8a55\u8ad6\u5167\u5bb9\u610f\u898b\u50be\u5411 \u8a08\u7b97\u65b9\u6cd5\u3002\u6839\u64da\u8a55\u8ad6\u4e3b\u984c\u4ee5\u53ca\u610f\u898b\u8a5e\u7684\u4fee\u98fe\u95dc\u4fc2\uff0c\u767c\u6398\u51fa\u4e3b\u984c\u76f8\u95dc\u7684\u8a55\u8ad6\u76ee\u6a19\u4ee5\u5224\u65b7\u5176 \u610f\u898b\u50be\u5411\u3002\u7136\u800c\u82e5\u8981\u63d0\u9ad8\u5206\u985e\u6e96\u78ba\u7387\uff0c\u9084\u9700\u8981\u9032\u4e00\u6b65\u627e\u51fa opinion holder \u8207 opinion polarity\uff0c \u751a\u81f3\u5728\u6642\u9593\u8ef8\u4e0a\u7684\u8b8a\u5316\u95dc\u4fc2, \u662f\u5c6c\u65bc\u6bd4\u8f03\u9ad8\u968e\u7684\u61c9\u7528\u3002\u9019\u6642\u5019\u6587\u53e5\u82e5\u80fd\u5148\u6a19 4. Bayes Classifier)\u4f86\u5206\u6790\u8a5e\u983b\u9032\u884c\u81ea\u52d5\u8a0a\u606f\u5206\u985e[18]\u3002\u6700\u5f8c\u6bd4\u8f03\u672c\u7814\u7a76\u65b9\u6cd5\u4ee5\u53ca\u50b3\u7d71\u5206\u985e \u8a5e\u6027\u3001\u6a19\u7247\u8a9e\u3001\u751a\u81f3\u5230\u53e5\u578b\u5256\u6790\u7b49\u524d\u8655\u7406\uff0c\u4e26\u64f7\u53d6\u8207\u4e3b\u984c\u76f8\u95dc\u7684\u5c6c\u6027\u4f86\u589e\u52a0\u627e\u51fa opinion</td></tr><tr><td>nowhere, nothing, nobody 0\u3002\u6700\u5f8c\u4f9d\u4e09\u7a2e\u5206\u6578\u7684\u500b\u5225\u52a0\u7e3d\uff0c\u63a1\u591a\u6578\u6c7a\u7684\u65b9\u5f0f\u4f86\u6c7a\u5b9a\u6bcf\u5247 tweet \u7684\u60c5\u7dd2\u9762\u5411\u3002 meaning(less) \u5982\u8868\u4e8c\u53ca\u8868\u4e09\u6240\u793a\uff0ctweet \u4e2d\u6240\u63d0\u53ca\u7684\u540d\u5b57\u53ca\u5c08\u6709\u540d\u8a5e\u4e0d\u4e00\u5b9a\u90fd\u548c\u96fb\u5f71\u76f8\u95dc\uff0c\u6240\u4ee5 \u70ba\"objective\"\u3002\u96d6\u7136\u6709\u627e\u5230\u96fb\u5f71\u540d\u5b57\u7684\u4fee\u98fe\u95dc\u4fc2\uff0c\u4f46 target \u4e26\u4e0d\u4e00\u5b9a\u8207\u8a72\u90e8\u96fb\u5f71\u76f8\u95dc\u3002 holder \u7684\u6e96\u78ba\u5ea6\uff0c\u4ee5\u4e0a\u90fd\u5c0d\u63d0\u9ad8\u60c5\u7dd2\u5206\u6790\u7684\u6e96\u78ba\u7387\u6703\u6709\u5e6b\u52a9\u3002</td></tr></table>", | |
| "type_str": "table", | |
| "num": null | |
| } | |
| } | |
| } | |
| } |