{ "paper_id": "O17-1006", "header": { "generated_with": "S2ORC 1.0.0", "date_generated": "2023-01-19T07:59:35.639161Z" }, "title": "Mining POIs from Web via POI recognition and Relation Verification", "authors": [ { "first": "Kuo-Hsin", "middle": [], "last": "\u8a31\u570b\u4fe1", "suffix": "", "affiliation": { "laboratory": "", "institution": "National Central University", "location": {} }, "email": "" }, { "first": "", "middle": [], "last": "Hsu", "suffix": "", "affiliation": { "laboratory": "", "institution": "National Central University", "location": {} }, "email": "" }, { "first": "Hsiu-Min", "middle": [], "last": "\u838a\u79c0\u654f", "suffix": "", "affiliation": { "laboratory": "", "institution": "National Central University", "location": {} }, "email": "" }, { "first": "", "middle": [], "last": "Chuang", "suffix": "", "affiliation": { "laboratory": "", "institution": "National Central University", "location": {} }, "email": "" }, { "first": "Chien-Lung", "middle": [], "last": "\u5468\u5efa\u9f8d", "suffix": "", "affiliation": { "laboratory": "", "institution": "National Central University", "location": {} }, "email": "" }, { "first": "", "middle": [], "last": "Chou", "suffix": "", "affiliation": { "laboratory": "", "institution": "National Central University", "location": {} }, "email": "" }, { "first": "Chia-Hui", "middle": [], "last": "\u5f35\u5609\u60e0", "suffix": "", "affiliation": { "laboratory": "", "institution": "National Central University", "location": {} }, "email": "" }, { "first": "", "middle": [], "last": "Chang", "suffix": "", "affiliation": { "laboratory": "", "institution": "National Central University", "location": {} }, "email": "" } ], "year": "", "venue": null, "identifiers": {}, "abstract": "This paper presents a system that could automatically extract new POIs from Web. First, we use special queries (e.g. Taipei+New Open) to find Web pages that might contain addresses for new stores. For web pages that contain addresses, we then apply store name recognition model to extract possible POIs. Finally, we train a model to find the most possible POI for the address found in the page. In this paper, we focus on POI name recognition and POI relation prediction. For POI recognition, we use store names from yellow pages as seed to prepare the training data via distant learning. Through entity selection and data processing, we obtain a model with 0.816 F1-measure as opposed to 0.432 F1-measure for a dictionary-based baseline. As for POI relation prediction, we compare three different strategies for negative example preparation. The best model could get 0.754 accuracy. We combine two POI recognition models with three classification models to test the overall performance. The best combination could extract 49 POIs every day with a single IP.", "pdf_parse": { "paper_id": "O17-1006", "_pdf_hash": "", "abstract": [ { "text": "This paper presents a system that could automatically extract new POIs from Web. First, we use special queries (e.g. Taipei+New Open) to find Web pages that might contain addresses for new stores. For web pages that contain addresses, we then apply store name recognition model to extract possible POIs. Finally, we train a model to find the most possible POI for the address found in the page. In this paper, we focus on POI name recognition and POI relation prediction. For POI recognition, we use store names from yellow pages as seed to prepare the training data via distant learning. Through entity selection and data processing, we obtain a model with 0.816 F1-measure as opposed to 0.432 F1-measure for a dictionary-based baseline. As for POI relation prediction, we compare three different strategies for negative example preparation. The best model could get 0.754 accuracy. We combine two POI recognition models with three classification models to test the overall performance. The best combination could extract 49 POIs every day with a single IP.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Abstract", "sec_num": null } ], "body_text": [], "back_matter": [], "bib_entries": { "BIBREF0": { "ref_id": "b0", "title": "Boosted Web Named Entity Recognition via Tri-Training", "authors": [ { "first": "Chien-Lung", "middle": [], "last": "Chou", "suffix": "" }, { "first": "Chia-Hui", "middle": [], "last": "Chang", "suffix": "" }, { "first": "Ya-Yun", "middle": [], "last": "Huang", "suffix": "" } ], "year": 2016, "venue": "ACM Trans. Asian Low-Resour. Lang. Inf. Process", "volume": "16", "issue": "10", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "Chien-Lung Chou, Chia-Hui Chang, and Ya-Yun Huang. 2016. Boosted Web Named Entity Recognition via Tri-Training. ACM Trans. Asian Low-Resour. Lang. Inf. Process. 16, 2, Article 10 (Oct. 2016), 23 pages.", "links": null }, "BIBREF1": { "ref_id": "b1", "title": "Verification of poi and location pairs via weakly labeled web data", "authors": [ { "first": "Hsiu-Min", "middle": [], "last": "Chuang", "suffix": "" }, { "first": "Chia-Hui", "middle": [], "last": "Chang", "suffix": "" } ], "year": 2015, "venue": "Proceedings of the 24th International Conference on World Wide Web", "volume": "", "issue": "", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "Chuang, Hsiu-Min, and Chia-Hui Chang. \"Verification of poi and location pairs via weakly labeled web data.\" Proceedings of the 24th International Conference on World Wide Web. ACM, 2015.", "links": null }, "BIBREF2": { "ref_id": "b2", "title": "A review of relation extraction", "authors": [ { "first": "Nguyen", "middle": [], "last": "Bach", "suffix": "" }, { "first": "Sameer", "middle": [], "last": "Badaskar", "suffix": "" } ], "year": 2007, "venue": "Literature review for Language and Statistics II", "volume": "", "issue": "", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "Bach, Nguyen, and Sameer Badaskar. \"A review of relation extraction.\" Literature review for Language and Statistics II (2007).", "links": null }, "BIBREF3": { "ref_id": "b3", "title": "Combining lexical, syntactic, and semantic features with maximum entropy models for extracting relations", "authors": [ { "first": "Nanda", "middle": [], "last": "Kambhatla", "suffix": "" } ], "year": 2004, "venue": "Proceedings of the ACL 2004 on Interactive poster and demonstration sessions", "volume": "", "issue": "", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "Kambhatla, Nanda. \"Combining lexical, syntactic, and semantic features with maximum entropy models for extracting relations.\" Proceedings of the ACL 2004 on Interactive poster and demonstration sessions. Association for Computational Linguistics, 2004.", "links": null }, "BIBREF4": { "ref_id": "b4", "title": "Extracting relations with integrated information using kernel methods", "authors": [ { "first": "Shubin", "middle": [], "last": "Zhao", "suffix": "" }, { "first": "Ralph", "middle": [], "last": "Grishman", "suffix": "" } ], "year": 2005, "venue": "Proceedings of the 43rd annual meeting on association for computational linguistics", "volume": "", "issue": "", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "Zhao, Shubin, and Ralph Grishman. \"Extracting relations with integrated information using kernel methods.\" Proceedings of the 43rd annual meeting on association for computational linguistics. Association for Computational Linguistics, 2005.", "links": null }, "BIBREF5": { "ref_id": "b5", "title": "International Workshop on The World Wide Web and Databases", "authors": [ { "first": "Sergey", "middle": [], "last": "Brin", "suffix": "" } ], "year": 1998, "venue": "", "volume": "", "issue": "", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "Brin, Sergey. \"Extracting patterns and relations from the world wide web.\" International Workshop on The World Wide Web and Databases. Springer, Berlin, Heidelberg, 1998.", "links": null }, "BIBREF6": { "ref_id": "b6", "title": "Open Information Extraction from the Web", "authors": [ { "first": "Michele", "middle": [], "last": "Banko", "suffix": "" } ], "year": 2007, "venue": "IJCAI", "volume": "7", "issue": "", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "Banko, Michele, et al. \"Open Information Extraction from the Web.\" IJCAI. Vol. 7. 2007.", "links": null }, "BIBREF7": { "ref_id": "b7", "title": "Unsupervised named-entity extraction from the web: An experimental study", "authors": [ { "first": "Oren", "middle": [], "last": "Etzioni", "suffix": "" } ], "year": 2005, "venue": "Artificial intelligence", "volume": "165", "issue": "", "pages": "91--134", "other_ids": {}, "num": null, "urls": [], "raw_text": "Etzioni, Oren, et al. \"Unsupervised named-entity extraction from the web: An experimental study.\" Artificial intelligence 165.1 (2005): 91-134.", "links": null }, "BIBREF8": { "ref_id": "b8", "title": "Snowball: Extracting relations from large plain-text collections", "authors": [ { "first": "Eugene", "middle": [], "last": "Agichtein", "suffix": "" }, { "first": "Luis", "middle": [], "last": "Gravano", "suffix": "" } ], "year": null, "venue": "Proceedings of the fifth ACM conference on Digital libraries", "volume": "", "issue": "", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "Agichtein, Eugene, and Luis Gravano. \"Snowball: Extracting relations from large plain-text collections.\" Proceedings of the fifth ACM conference on Digital libraries.", "links": null }, "BIBREF9": { "ref_id": "b9", "title": "An algorithm that learns what's in a name", "authors": [ { "first": "Daniel", "middle": [ "M" ], "last": "Bikel", "suffix": "" }, { "first": "Richard", "middle": [], "last": "Schwartz", "suffix": "" }, { "first": "Ralph", "middle": [ "M" ], "last": "Weischedel", "suffix": "" } ], "year": 1999, "venue": "Machine learning", "volume": "34", "issue": "", "pages": "211--231", "other_ids": {}, "num": null, "urls": [], "raw_text": "Bikel, Daniel M., Richard Schwartz, and Ralph M. Weischedel. \"An algorithm that learns what's in a name.\" Machine learning 34.1 (1999): 211-231.", "links": null }, "BIBREF10": { "ref_id": "b10", "title": "Early results for named entity recognition with conditional random fields, feature induction and web-enhanced lexicons", "authors": [ { "first": "Andrew", "middle": [], "last": "Mccallum", "suffix": "" }, { "first": "Wei", "middle": [], "last": "Li", "suffix": "" } ], "year": 2003, "venue": "Proceedings of the seventh conference on Natural language learning at HLT-NAACL 2003", "volume": "4", "issue": "", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "McCallum, Andrew, and Wei Li. \"Early results for named entity recognition with conditional random fields, feature induction and web-enhanced lexicons.\" Proceedings of the seventh conference on Natural language learning at HLT-NAACL 2003-Volume 4. Association for Computational Linguistics, 2003.", "links": null } }, "ref_entries": { "TABREF0": { "text": "SR \u806f\u96c6\u5f8c\u53d6\u5f97 556,036 \u500b\u5be6\u9ad4\uff0c\u900f\u904e\u641c\u5c0b\u70ba\u6bcf\u500b\u5be6\u9ad4\u53d6\u5f97\u76f8\u95dc\u53e5\u5b50\u505a\u70ba\u8a13 \u7df4\u8cc7\u6599\uff0c\u7d93\u8655\u7406\u5f8c\u5171\u6709 928,567 \u53e5\u3002\u800c\u5229\u7528 L \u548c SP \u806f\u96c6\u5f8c\u5247\u6709 556,702 \u500b\u5be6\u9ad4\uff0c\u4ee5\u53ca 916,383 \u8a13\u7df4\u53e5\u3002\u800c\u539f\u5148 677,172 \u500b\u9ec3\u9801\u8208\u8da3\u9ede\uff0c\u5247\u6709 1,560,622 \u5305\u542b\u5be6\u9ad4\u7684\u53e5\u5b50\u3002 3.2 \u81ea\u52d5\u6a19\u8a18 \u63a1\u7528\u5df2\u77e5\u5be6\u9ad4\u540d\u7a31\uff0c\u81ea\u52d5\u6a19\u8a18\u53e5\u5b50\u751f\u6210\u8a13\u7df4\u8cc7\u6599\uff0c\u662f\u89e3\u6c7a\u4eba\u5de5\u6a19\u8a18\u6210\u672c\u904e\u65bc\u6602\u8cb4\u7684\u65b9\u6cd5\u3002 \u96d6\u7136\u53e5\u5b50\u662f\u7531\u67e5\u8a62\u5df2\u77e5\u8208\u8da3\u9ede\u53d6\u5f97\uff0c\u4f46\u7531\u65bc\u6bcf\u53e5\u8a71\u53ef\u80fd\u5305\u542b\u4e00\u500b\u6216\u4e00\u500b\u4ee5\u4e0a\u7684\u8208\u8da3\u9ede\uff0c \u4e5f\u53ef\u80fd\u5b8c\u5168\u6c92\u6709(\u5982\u5716\u4e09) \u3002\u7576\u53e5\u5b50\u6578\u91cf\u8207\u5be6\u9ad4\u540d\u7a31\u6578\u91cf\u5747\u5927\u65bc\u5e7e\u5341\u842c\u6642\uff0c\u81ea\u52d5\u6a19\u8a18\u7684 \u6210\u672c\u5c31\u6703\u76f8\u7576\u5927\u3002\u70ba\u907f\u514d\u5de2\u72c0\u6a19\u8a18(\u5373\u4e00\u500b POI \u88e1\u5305\u542b\u53e6\u4e00\u500b POI)\u8207\u52a0\u901f\u6a19\u8a18\u901f\u5ea6\uff0c\u6211", "html": null, "content": "
\u4e00\u3001\u7dd2\u8ad6 \u96a8\u8457\u7121\u7dda\u7db2\u8def\u548c\u667a\u6167\u578b\u624b\u6a5f\u7684\u666e\u53ca\uff0c\u50b3\u7d71\u7ffb\u95b1\u96fb\u8a71\u7c3f\u6216\u540d\u7247\u7684\u65b9\u5f0f\u5927\u5e45\u6e1b\u5c11\uff0c\u4f7f\u7528\u8005\u5011 \u958b\u59cb\u7fd2\u6163\u5229\u7528\u7db2\u8def\u67e5\u8a62\u5e97\u5bb6\u8cc7\u8a0a\u3002\u56e0\u6b64\uff0c\u8c37\u6b4c\u3001\u96c5\u864e\u3001\u5fae\u8edf\u548c\u8afe\u57fa\u4e9e\u7b49\u516c\u53f8\u5f88\u65e9\u5c31\u5df2\u958b \u59cb\u958b\u767c\u5546\u696d\u5730\u5716\u4ee5\u6eff\u8db3\u9019\u985e\u7684\u9700\u6c42\uff0c\u4e5f\u6709\u5176\u4ed6\u516c\u5171\u7684\u5730\u5716\uff0c\u50cf\u662f\uff1aOpenStreetMap \u548c OpenPOI\uff0c\u7528\u65bc\u5efa\u8a2d\u548c\u7dad\u8b77\u8208\u8da3\u9ede(Point of Interest, POI)\u6578\u64da\u5eab\u3002 \u5728 W3C \u7684\u5b9a\u7fa9\u4e2d\uff0c\u8208\u8da3\u9ede\u53ef\u4ee5\u8996\u70ba\u4e00\u500b\u64c1\u6709\u53ef\u7528\u4fe1\u606f\u7684\u4f4d\u7f6e\u3002\u5ee3\u7fa9\u4e0a\u4f86\u8aaa\uff0c\u4efb\u4f55\u80fd\u5728 \u96fb\u5b50\u5730\u5716\u4e0a\u6a19\u8a18\u7684\u67d0\u500b\u5730\u6a19\u6216\u662f\u5efa\u7bc9\uff0c\u90fd\u53ef\u4ee5\u7576\u4f5c\u8208\u8da3\u9ede\uff0c\u50cf\u662f\uff1a\u4e2d\u592e\u5927\u5b78\u3001\u6377\u904b\u4e2d\u5c71 \u7ad9\u3001\u53f0\u5317 101 \u7b49\u7b49\u3002 \u5730\u5716\u670d\u52d9\u4e0d\u50c5\u50c5\u5305\u542b\u9910\u98f2\u5e97\u5bb6\uff0c\u5c1a\u6709\u65c5\u904a\u3001\u91ab\u7642\u3001\u5b78\u6821\u3001\u4e94\u91d1\u3001\u6c34\u96fb\u7b49\u5176\u4ed6\u5e97\u5bb6\uff0c\u5e73\u5747 \u6bcf\u5929\u65b0\u589e\u7684\u8208\u8da3\u9ede\u53ef\u80fd\u6709\u4e0a\u767e\u500b\u3002\u4e0d\u53ef\u80fd\u53ea\u501a\u9760\u771f\u4eba\u8003\u5bdf\u6216\u700f\u89bd\u90e8\u843d\u5ba2\u6587\u7ae0\u53bb\u65b0\u589e\u8208\u8da3 \u9ede\uff0c\u56e0\u6b64\u9700\u8981\u81ea\u52d5\u5316\u3002 \u7531\u65bc\u6211\u5011\u53ef\u4ee5\u5229\u7528\u5730\u5740\u95dc\u9375\u5b57(\u5340\u3001\u5e02\u3001\u93ae)\u505a\u70ba\u641c\u5c0b\u5b57\u53bb\u6293\u53d6\u7db2\u9801\u4e2d\u7684\u5730\u5740\uff0c\u56e0\u6b64\u672c\u8ad6 \u6587\u4ee5\u5730\u5740\u7684\u89d2\u5ea6\u53bb\u64ec\u5b9a\u7b56\u7565\uff0c\u5e0c\u671b\u5feb\u901f\u5730\u5f9e\u7db2\u8def\u4e0a\u6316\u6398\u51fa\u65b0\u7684\u5730\u5740\uff0c\u4e26\u7d66\u4e88\u6bcf\u500b\u65b0\u7684\u5730 \u5740\u6b63\u78ba\u7684\u8208\u8da3\u9ede\u540d\u7a31\uff0c\u4ee5\u6b64\u81ea\u52d5\u5316\u64f4\u5145\u8cc7\u6599\u5eab\u3002\u6211\u5011\u63d0\u51fa\u7684\u7cfb\u7d71\u5305\u542b\u56db\u500b\u90e8\u5206\uff1a\u7b2c\u4e00\u90e8 \u5206\u70ba\u95dc\u9375\u5b57\u722c\u87f2\uff0c\u7b2c\u4e8c\u90e8\u5206\u70ba\u5730\u5740\u8fa8\u8b58\uff0c\u7b2c\u4e09\u90e8\u5206\u70ba\u8208\u8da3\u9ede\u8fa8\u8b58\uff0c\u6700\u5f8c\u5247\u662f\u914d\u5c0d\u95dc\u4fc2\u9810 \u6e2c\uff0c\u5982\u5716\u4e00\u6240\u793a\u3002\u7cfb\u7d71\u53ef\u4ee5\u900f\u904e\u7279\u5b9a\u7684\u95dc\u9375\u5b57(ex:\u53f0\u5317\u5e02+\u65b0\u958b\u5e55)\u9032\u884c\u7b2c\u4e00\u6b21\u8cc7\u6599\u8490\u96c6\uff0c \u5f9e\u4e2d\u64f7\u53d6\u51fa\u8fa8\u8b58\u7684\u5730\u5740\uff0c\u4ee5\u53ca\u5305\u542b\u5730\u5740\u7db2\u9801\u4e2d\u7684\u8208\u8da3\u9ede\uff0c\u6700\u5f8c\u5c0d\u6bcf\u500b\u5730\u5740\u548c\u5176\u627e\u5230\u7684\u8208 \u8da3\u9ede\u505a\u95dc\u4fc2\u9810\u6e2c\uff0c\u9078\u64c7\u6a5f\u7387\u6700\u9ad8\u7684\u505a\u70ba\u6b63\u78ba\u914d\u5c0d\u3002 \u5716\u4e00\u3001\u7cfb\u7d71\u67b6\u69cb\u5716 \u96d6\u7136\u5c07 Google \u641c\u5c0b\u5f15\u64ce\u4f5c\u70ba\u5927\u91cf\u8cc7\u8a0a\u7684\u4f86\u6e90\uff0c\u6293\u53d6\u8a72\u95dc\u9375\u5b57\u7684\u524d\u5341\u7bc7\u641c\u5c0b\u7d50\u679c\u53ef\u4ee5\u5feb \u901f\u7684\u627e\u5c0b\u65b0\u7684\u5546\u5bb6\uff0c\u4e0d\u904e\u9019\u9805\u641c\u5c0b\u4f86\u6e90\u7684\u9650\u5236\u662f\u540c IP \u4e0d\u80fd\u983b\u7e41\u5730\u5411 Google \u641c\u5c0b\u5f15\u64ce\u8490 \u96c6\u8cc7\u6599\uff0c\u56e0\u6b64\u7cfb\u7d71\u6bcf\u5929\u80fd\u81ea\u52d5\u5316\u627e\u5230\u591a\u5c11\u500b\u65b0\u7684\u5546\u5bb6\u662f\u5be6\u4f5c\u5730\u7406\u8cc7\u8a0a\u7cfb\u7d71\u6240\u95dc\u5fc3\u7684\u4e3b\u984c\u3002 \u672c\u7bc7\u8ad6\u63a1\u7528 Huang [12]\u4e4b\u65b9\u6cd5\u64f7\u53d6\u4e2d\u6587\u5730\u5740(F1 \u503c\u53ef\u9054 97.2%)\uff0c\u4e26\u6539\u5584\u8208\u8da3\u9ede\u8fa8\u8b58\u9054\u5230 \u4e8c\u3001\u76f8\u95dc\u7814\u7a76 \u6b64\u9700\u8981\u9032\u4e00\u6b65\u7684\u7be9\u9078\uff0c\u904e\u6ffe\u6389\u4e0d\u7b26\u5408\u8208\u8da3\u9ede\u5b9a\u7fa9\u7684\u5be6\u9ad4\u3002 \u8868\u4e00\u3001\u8208\u8da3\u9ede\u8fa8\u8b58\u6a21\u578b\u7279\u5fb5\u8868 ID Name Description 1 Before_1 unigram word before entity or not \u7279\u5fb5\u4e00\u5230\u7279\u5fb5\u4e09\u5229\u7528\u641c\u5c0b\u7d50\u679c\u6578\u53d6\u5c0d\u6578\u800c\u5f97\u3002\u6211\u5011\u8a8d\u70ba\u5730\u5740\u6216\u8208\u8da3\u9ede\u7684\u641c\u5c0b\u7d50\u679c\u6578\u8d8a\u4f4e\uff0c \u4ee3\u8868\u5176\u4e0d\u5b58\u5728\u7684\u6a5f\u7387\u8d8a\u9ad8\uff0c\u5728\u9a57\u8b49\u6642\u88ab\u5206\u985e\u70ba\u932f\u8aa4\u914d\u5c0d(False)\u7684\u6a5f\u7387\u5c31\u6703\u8d8a\u9ad8\u3002\u4ee5\u5730\u5740 \u8868\u4e09\u3001\u6e2c\u8a66\u8cc7\u6599\u4e4b\u4e00\u81f4\u6027\u4fe1\u5ea6\u503c Labeler1 \u5373\u53ef\u770b\u51fa\uff0c\u5229\u7528\u96a8\u6a5f\u5206\u914d\u7522\u751f\u53cd\u4f8b\u7684\u95dc\u806f\u5206\u985e\u6a21\u578b\u80fd\u627e\u5230\u8f03\u591a\u7684 POI\uff0c\u6578\u91cf\u70ba\u5176\u4ed6\u5169\u500b 0.744 \u548c 0.754\uff0c\u5dee\u7570\u4e26\u4e0d\u986f\u8457\u3002 5.2 \u5730\u5740\u8207\u8208\u8da3\u9ede\u95dc\u806f\u9810\u6e2c \u7684\u5e7e\u5341\u500d\u3002\u9019\u80cc\u5f8c\u7684\u539f\u56e0\u53ef\u80fd\u662f\u56e0\u70ba\u6b63\u4f8b\u548c\u53cd\u4f8b\u7684\u5dee\u7570\u8f03\u5927\uff0c\u80fd\u8f15\u6613\u5c07\u6b63\u78ba\u7b54\u6848\u8207\u932f\u8aa4 \u6574\u9ad4\u7cfb\u7d71\u65b9\u9762\uff0c\u6211\u5011\u63a1\u53d6\u516d\u7a2e\u4e0d\u540c\u7684\u7d44\u5408\u5c0d\u4e00\u767e\u500b\u65b0\u7684\u5730\u5740\u9032\u884c\u6e2c\u8a66\uff0c\u5f9e\u7d50\u679c\u4e2d\u53ef\u4ee5\u770b \u5be6\u9ad4\u63d0\u53d6\u662f\u5f9e\u975e\u7d50\u69cb\u5316\u6587\u672c\u6587\u6a94\u4e2d\u8b58\u5225\u547d\u540d\u5be6\u9ad4\u7684\u4efb\u52d9\uff0c\u9019\u662f\u7528\u65bc\u6e2c\u8a66\u6a5f\u5668\u80fd\u5920\u7406\u89e3\u81ea \u7136\u8a9e\u8a00\u5beb\u5165\u7684\u6d88\u606f\u4ee5\u53ca\u81ea\u52d5\u57f7\u884c\u901a\u5e38\u57f7\u884c\u7684\u5e38\u898f\u4efb\u52d9\u7684\u4fe1\u606f\u4efb\u52d9\u4e4b\u4e00\u3002\u73fe\u4eca\u5e38\u7528\u7684\u65b9\u5f0f \u662f\u4ee5\u5e8f\u5217\u6a19\u8a18\u5be6\u9ad4\u7684\u958b\u59cb\u3001\u4e2d\u7e7c\u3001\u7d50\u675f\u3001\u5176\u4ed6\u4f5c\u70ba\u64f7\u53d6\u7684\u53c3\u7167\uff0c\u4e26\u4ee5\u96b1\u85cf\u99ac\u723e\u53ef\u592b\u6a21\u578b (Hidden Markov Model, HMM)\u548c\u689d\u4ef6\u96a8\u6a5f\u5834(Conditional Random Field, CRF)\u70ba\u4e3b \u8981\u6280\u8853[11]\u3002\u7531\u65bc\u76e3\u7763\u5b78\u7fd2\u9700\u6e96\u5099\u5927\u91cf\u7684\u8a13\u7df4\u8cc7\u6599\uff0c\u800c\u4eba\u5de5\u6a19\u8a18\u9700\u8981\u76f8\u95dc\u77e5\u8b58\u4e14\u8017\u8cbb\u6642 \u9593\u3002\u56e0\u6b64 Chou[1]\u8207 Huang[12]\u7b49\u4eba\u5373\u63d0\u51fa\u5229\u7528\u5df2\u77e5\u7684\u5be6\u9ad4\u6e05\u55ae\u9032\u884c\u81ea\u52d5\u6a19\u8a18\u9032\u800c\u751f\u6210 \u8a13\u7df4\u8cc7\u6599\u7684 Distant Learning \u67b6\u69cb\u3002\u672c\u8ad6\u6587\u4e2d\u7684\u8208\u8da3\u9ede\u8fa8\u8b58\u6a21\u578b\u5373\u4eff\u6548\u9ec3\u7684\u505a\u6cd5\uff0c\u4e26\u6539 \u5584\u8fa8\u8b58\u6548\u80fd\uff0c\u6b64\u70ba\u672c\u7814\u7a76\u7684\u7b2c\u4e00\u500b\u4e3b\u984c\u3002 \u672c\u6587\u7b2c\u4e8c\u500b\u4e3b\u984c\u5247\u662f\u5730\u5740\u8207\u8208\u8da3\u9ede\u95dc\u4fc2\u7684\u9a57\u8b49\u6a21\u578b\u3002\u63d0\u53d6\u5be6\u9ad4\u4e4b\u9593\u7684\u8a9e\u7fa9\u95dc\u4fc2\u662f\u6578\u64da\u93c8 \u63a5\u8207\u672c\u9ad4\u5efa\u69cb\u767c\u5c55\u7684\u95dc\u9375\u6b65\u9a5f\u3002\u5927\u591a\u6578\u7684\u7814\u7a76\u8457\u91cd\u5728\u4e8c\u5143\u95dc\u4fc2\u7684\u64f7\u53d6[3]\uff0c\u67d0\u4e9b\u76e3\u7763\u5f0f \u5b78\u7fd2\u7684\u7814\u7a76\u63d0\u51fa\u7279\u5fb5\u5c0e\u5411[4]\u4ee5\u53ca\u6838\u5fc3(kernel-based)\u5c0e\u5411[5]\u7684\u65b9\u6cd5\u3002\u7531\u65bc\u76e3\u7763\u5f0f\u5b78\u7fd2\u9700\u8981 \u5927\u91cf\u6a19\u8a18\u8cc7\u6599\uff0c\u56e0\u6b64\u534a\u76e3\u7763\u5f0f\u5b78\u7fd2(Semi-supervised Learning)\u548c\u81ea\u52a9\u6cd5(bootstrapping)\u5c31 \u986f\u5f97\u91cd\u8981\uff0cDIPRE [6]\u548c Snowball [9]\u5206\u5225\u4f7f\u7528\u4e00\u5c0f\u7d44\u6a19\u8a18\u7a2e\u5b50\u5be6\u4f8b\u548c\u624b\u5de5\u63d0\u53d6\u683c\u5f0f\u4f86\u8a13 \u7df4\u6a21\u578b\u3002\u800c KnowItAll[8]\u548c TextRunner[7]\u5247\u662f\u63a1\u7528\u81ea\u6211\u8a13\u7df4\u4e4b\u5927\u578b\u95dc\u4fc2\u64f7\u53d6\u7cfb\u7d71\u3002 \u672c\u6587\u63a1\u7528\u9ad8\u9706\u8000\u53ca\u838a\u79c0\u654f[13]\u7b49\u4eba\u7684\u4f5c\u6cd5\uff0c\u900f\u904e\u641c\u5c0b\u5730\u5740\u8207\u5e97\u5bb6\u7684\u7d50\u679c\u6578\u3001\u76ae\u723e\u68ee\u76f8\u95dc \u4fc2\u6578 (Pearson correlation coefficient)\u3001\u9918\u5f26\u76f8\u4f3c\u5ea6(Cosine similarity)\u7b49\u5171 27 \u500b\u7279\u5fb5\u53bb\u63a8 \u65b7\u8a72\u5546\u5bb6\u662f\u5426\u4f4d\u5728\u8a72\u5730\u5740\u4e0a\uff0c\u4e26\u85c9\u7531\u9ec3\u9801\u7684\u5546\u5bb6\u8cc7\u8a0a\u6e96\u5099\u6b63\u53cd\u914d\u5c0d\u8a13\u7df4\u53ca\u6e2c\u8a66\u8cc7\u6599\uff0c\u7136 \u5be6\u52d9\u4e0a\u5730\u5740\u8207\u5546\u5bb6\u914d\u5c0d\u4e4b\u6e2c\u8a66\u8cc7\u6599\u8207\u8a13\u7df4\u8cc7\u6599\u4e26\u4e0d\u76f8\u540c\uff0c\u70ba\u52a0\u901f\u7cfb\u7d71\u904b\u4f5c\uff0c\u672c\u6587\u63d0\u51fa\u65b0 \u7684\u8a13\u7df4\u8cc7\u6599\u6e96\u5099\u65b9\u5f0f\uff0c\u5e0c\u671b\u53ef\u4ee5\u63d0\u5347\u7cfb\u7d71\u904b\u4f5c\u6548\u80fd\u3002 \u4e09\u3001\u8208\u8da3\u9ede\u8fa8\u8b58\u6a21\u7d44 \u8208\u8da3\u9ede\u8fa8\u8b58\u6a21\u7d44\u5305\u542b\u4e94\u500b\u6b65\u9a5f\uff0c\u5305\u62ec\u4ee5\u5df2\u77e5\u8208\u8da3\u9ede\u4f5c\u70ba\u95dc\u9375\u5b57\u67e5\u8a62\u53ef\u80fd\u5305\u542b\u8208\u8da3\u9ede\u7684\u53e5 \u5716\u4e8c\u3001\u8208\u8da3\u9ede\u8fa8\u8b58\u6a21\u7d44\u6d41\u7a0b\u5716 3.1 \u5546\u5bb6\u5be6\u9ad4\u7be9\u9078 \u7531\u65bc\u9ec3\u9801\u4e2d\u7684\u8208\u8da3\u9ede\u53ef\u80fd\u5305\u542b\u82f1\u6587\u3001\u6578\u5b57\u6216\u7279\u6b8a\u7b26\u865f\uff0c\u800c\u6211\u5011\u7684\u7814\u7a76\u8457\u91cd\u5728\u7d14\u4e2d\u6587\u7684\u8208 \u8da3\u9ede\u540d\u7a31\uff0c\u56e0\u6b64\u4fdd\u7559\u7531\u4e2d\u6587\u4ee5\u53ca\u62ec\u865f\u7d44\u6210\u7684\u5be6\u9ad4\uff0c\u5176\u9918\u5168\u90e8\u53bb\u9664\u3002\u4fdd\u7559\u62ec\u865f\u662f\u56e0\u70ba\u67d0\u4e9b \u8208\u8da3\u9ede\u6703\u8ddf\u96a8\u8457\u5340\u57df\u540d\uff0c\u50cf\u662f\u9023\u9396\u5e97\u5c31\u6703\u900f\u904e\u4e0d\u540c\u9580\u5e02\u4f86\u5340\u5206\uff0c\u8209\u4f8b\u4f86\u8aaa\u300c\u5168\u5bb6(\u4e2d\u592e \u5e97)\u300d\u548c\u300c\u5168\u5bb6(\u4e2d\u6b63\u5e97)\u300d\u540c\u6a23\u90fd\u662f\u4fbf\u5229\u8d85\u5546\uff0c\u4f46\u7531\u65bc\u843d\u5728\u4e0d\u540c\u5340\u57df\u4e5f\u5c31\u6703\u6709\u4e0d\u540c\u7684\u9580\u5e02 \u540d\u7a31\u3002 \u7be9\u9078\u7684\u898f\u5247\u53ef\u4ee5\u4f9d\u64da\u5be6\u9ad4\u7684\u9577\u77ed\u5206\u6210\u5169\u90e8\u5206(\u9577\u5ea6\u70ba\u4e94\u5230\u5341\u4e94\u70ba\u9577\u5be6\u9ad4\uff1b\u9577\u5ea6\u4e09\u6216\u56db\u5247 \u662f\u77ed\u5be6\u9ad4)\uff0c\u5176\u4e2d\u91dd\u5c0d\u77ed\u5be6\u9ad4\u53ef\u80fd\u70ba\u4eba\u540d\u7684\u90e8\u5206\u518d\u52a0\u4ee5\u7d30\u5206\u5169\u7a2e\u898f\u5247\u3002\u4e4b\u5f8c\u4f7f\u7528 1,563 \u500b\u98df\u7269\u540d\u7a31\u4ee5\u53ca 1,275 \u500b\u985e\u5225\u540d\u7a31\u9032\u884c\u904e\u6ffe\uff0c\u4e26\u5229\u7528\u6b63\u898f\u8868\u9054\u5f0f(\u5340\u9109\u93ae\u7e23\u5e02\u90e8)\u53bb\u9664\u5730 \u5340\u540d\u7a31\uff0c\u907f\u514d\u64f7\u53d6\u51fa\u5927\u7bc4\u570d\u7684\u5730\u9ede\u540d\u7a31\uff0c\u7559\u4e0b\u9577\u5be6\u9ad4\u7684\u8208\u8da3\u9ede\uff0c\u4ee5\u4e0b\u7c21\u7a31\u6b64\u6e05\u55ae\u70ba L\u3002 \u63a5\u8457\uff0c\u6211\u5011\u900f\u904e\u6b63\u898f\u8868\u9054\u5f0f\u4ee5\u53ca\u4eba\u540d\u8fa8\u8b58\u6a21\u578b\u904e\u6ffe\u8a3b\u518a\u4eba\u540d\u3002\u6b63\u898f\u8868\u9054\u5f0f\u7684\u65b9\u6cd5\u4fc2\u5229\u7528 124 \u500b\u5e38\u7528\u59d3\u6c0f\u505a\u70ba\u958b\u982d\uff0c\u7372\u5f97\u53bb\u9664\u53ef\u80fd\u4eba\u540d\u5f8c\u7684\u77ed\u5be6\u9ad4\u8208\u8da3\u9ede\uff0c\u4ee5\u4e0b\u7c21\u7a31\u6b64\u6e05\u55ae SR\u3002 \u4ee5\u4eba\u540d\u8fa8\u8b58\u6a21\u578b\u7684\u90e8\u5206\uff0c\u5247\u5148\u4ee5\u77ed\u5be6\u9ad4\u505a\u70ba\u641c\u5c0b\u95dc\u9375\u5b57\u722c\u53d6\u524d\u5341\u7bc7\u641c\u5c0b\u7d50\u679c\uff0c\u4e26\u5229\u7528\u4eba \u540d\u8fa8\u8b58\u6a21\u578b\u64f7\u53d6\u4eba\u540d\u6e05\u55ae[12]\uff0c\u548c\u539f\u5148\u7684\u77ed\u5be6\u9ad4\u6e05\u55ae\u53d6\u5dee\u96c6\u5f8c\u5373\u70ba\u904e\u6ffe\u5b8c\u6210\u7684\u77ed\u5be6\u9ad4\u6e05 \u55ae\uff0c\u4ee5\u4e0b\u7c21\u7a31\u6b64\u6e05\u55ae SP\u3002 \u6c92\u6709\u660e\u986f\u4e0b\u964d\uff0c\u5be6\u9a57\u6642\u9593\u5247\u6709\u5927\u5e45\u5ea6\u7684\u964d\u4f4e\u3002 \u6211\u5011\u4f9d\u64da Chou[1]\u63d0\u51fa\u7684\u4e94\u985e\u5341\u56db\u7a2e\u7279\u5fb5(\u5982\u8868\u4e00)\uff1a\u662f\u5426\u70ba\u5be6\u9ad4\u524d\u5e38\u51fa\u73fe\u7684\u8a5e(Common Before)\u3001\u662f\u5426\u70ba\u5be6\u9ad4\u5f8c\u5e38\u51fa\u73fe\u7684\u8a5e(Common After)\u3001\u5e38\u51fa\u73fe\u7684\u5be6\u9ad4\u524d\u7db4\u8a5e\u662f(Common Prefix)\u3001\u5e38\u51fa\u73fe\u7684\u5be6\u9ad4\u5f8c\u7db4\u8a5e(Common Postfix)\uff0c\u4ee5\u53ca\u662f\u5426\u70ba\u7279\u6b8a(\u5982\u82f1\u6587\u3001\u6578\u5b57\u7b49)\u7b26\u865f\u3002 \u5730\u8fa8\u8b58\u51fa\u5730\u5740\u548c\u8208\u8da3\u9ede\u95dc\u806f\uff0c\u6211\u5011\u5b9a\u7fa9\u4ee5\u4e0b\u5982\u8868\u4e8c\u3001\u5171\u5341\u4e8c\u500b\u7279\u5fb5\u3002 \u8868\u4e8c\u3001\u5730\u5740\u8207\u8208\u8da3\u9ede\u95dc\u806f\u5206\u985e\u7279\u5fb5\u8868 \u672c\u7bc0\u5167\u5bb9\u5305\u542b\u4e09\u500b\u90e8\u4efd\uff0c\u5206\u5225\u662f\u8208\u8da3\u9ede\u8fa8\u8b58\u6548\u80fd\u7684\u6bd4\u8f03\u3001\u5730\u5740\u8207\u5546\u5bb6\u95dc\u4fc2\u9810\u6e2c\u3001\u4ee5\u53ca\u6574 \u9ad4\u7cfb\u7d71\u6548\u80fd\u3002 5.1 \u8208\u8da3\u9ede\u8fa8\u8b58 \u5f9e\u8868\u56db\u4e2d\u53ef\u4ee5\u770b\u51fa\uff0c\u5229\u7528 LSP10 \u642d\u914d\u7b2c\u4e00\u7a2e\u95dc\u806f\u5206\u985e\u6a21\u578b\u7684\u6548\u80fd\u6700\u597d\uff0c\u6e96\u78ba\u7387\u9054\u5230 0.648\uff0c 37.3%\u7684\u8a13\u7df4\u6642\u9593\uff0c\u5176\u6548\u80fd\u548c\u4fdd\u7559\u6240\u6709\u53e5\u5b50\u7684\u65b9\u5f0f\u4e0d\u76f8\u4e0a\u4e0b\u3002\u800c\u53bb\u9664\u4e0d\u5305\u542b\u5be6\u9ad4\u4ee5\u53ca\u6a19 \u8208\u8da3\u9ede\u8fa8\u8b58\u8a55\u4f30\u65b9\u5f0f\u63a1\u53d6\u90e8\u5206\u6bd4\u5c0d\uff0c\u5047\u5982\u6a19\u6e96\u7b54\u6848\u662f\u300c\u8523\u4e2d\u6b63\u7d00\u5ff5\u9928\u300d \uff0c\u800c\u6a21\u578b\u53ea\u8fa8\u8b58 \u4e5f\u53ea\u6709 0.432\u3002\u800c\u7d93\u904e\u8a13\u7df4\u7684\u6a21\u578b\uff0c\u5373\u4f7f\u662f\u6c92\u6709\u7d93\u904e\u7be9\u9078(ALL)\u7684\u6a21\u578b F1 \u6548\u80fd\u4e5f\u6709 0.665\uff1b \u8868\u73fe\u6700\u597d\u7684\u662f\u904e\u6ffe\u4eba\u540d\u7684 LSP \u8fa8\u8b58\u6a21\u578b\uff0cF1 \u6548\u80fd\u70ba 0.788\uff0c\u800c \u6b63\u898f\u8868\u9054\u5f0f\u4eba\u540d\u904e\u6ffe\u6240\u8a13 \u7df4\u7684\u6a21\u578b\u6548\u80fd\u5247\u9054\u5230 0.770\u3002 \u5728\u6548\u7387\u4e0a\uff0c\u6bd4\u8f03\u4fdd\u7559\u6240\u6709\u641c\u5c0b\u7d50\u679c\u7684\u5168\u90e8\u53e5\u5b50 (2,426,201 \u53e5)\u6240\u8a13\u7df4\u7684\u6a21\u578b\uff0c\u8207\u53bb\u9664\u4e0d \u5305\u542b\u8208\u8da3\u9ede\u7684\u53e5\u5b50(916,383 \u53e5)\u6240\u8a13\u7df4\u7684\u6a21\u578b\u3002\u5f9e\u5716\u5341\u4e8c\u53ef\u770b\u51fa\uff0c\u53bb\u9664\u4e0d\u5305\u542b\u8208\u8da3\u9ede\u7684 \u6a21\u578b\u548c\u4fdd\u7559\u7684\u6a21\u578b\u6548\u80fd\u76f8\u5dee\u4e0d\u9060\uff0c\u5be6\u9a57\u6642\u9593\u537b\u5f9e 757 \u5206\u9418\u964d\u5230 474 \u5206\u9418\uff0c\u5171\u6e1b\u5c11 37.3% \u7684\u6642\u9593\u3002\u7531\u6b64\u53ef\u77e5\uff0c\u53bb\u9664\u4e0d\u5305\u542b\u8208\u8da3\u9ede\u80fd\u5927\u5e45\u964d\u4f4e\u8a13\u7df4\u6642\u9593\uff0c\u4e14\u6548\u80fd\u4e0d\u6703\u6709\u592a\u591a\u7684\u8b8a\u52d5\u3002 \u65bc 0.5 \u7684\u5019\u9078\u8208\u8da3\u9ede(\u8868\u56db\u7b2c 3 \u884c)\u4e26\u9032\u884c\u6392\u5e8f\uff0c\u9078\u51fa\u6700\u9ad8\u6a5f\u7387\u7684\u5019\u9078\u8005\u505a\u70ba\u8a72\u5730\u5740\u914d\u5c0d \u7684\u8208\u8da3\u9ede\uff0c\u6700\u7d42\u4eba\u5de5\u6a19\u8a18\u7d50\u679c\u4e26\u8a08\u7b97\u6e96\u78ba\u7387(\u8868\u56db\u7b2c 4 \u884c)\u3002\u5176\u4e2d\u5229\u7528 ALL \u6a21\u578b\u8fa8\u8b58\u51fa \u7684\u8208\u8da3\u9ede\u5171\u6709 816 \u500b\uff0cLSP10 \u5247\u8fa8\u8b58\u51fa 340 \u500b\u8208\u8da3\u9ede\uff0c\u56e0\u6b64\u524d\u4e09\u7d44\u6240\u9700\u8981\u7684\u9a57\u8b49\u6642\u9593\u4e5f \u662f\u5f8c\u8005\u7684 2.4 \u500d\uff0c\u518d\u7531\u6b64\u82b1\u8cbb\u7684\u6642\u9593\u63a8\u7b97\u6bcf\u65e5\u53ef\u4ee5\u627e\u5230\u65b0(\u6b63\u78ba)\u7684\u5730\u5740\u5546\u5bb6\u914d\u5c0d\u3002 \u7684\u8208\u8da3\u9ede\u540d\u7a31\u5305\u542b\u4e0d\u53ea\u4e00\u822c\u5546\u5bb6\u540d\u7a31\uff0c\u4e5f\u5305\u62ec\u50cf\u662f\u8a3b\u518a\u4eba\u540d\u3001\u98df\u7269\u540d\u6216\u662f\u985e\u5225\u540d\u7a31\u7684\u8208 \u8da3\u9ede\uff0c\u6240\u4ee5\u9700\u8981\u9032\u884c\u8208\u8da3\u9ede\u7684\u7be9\u9078\uff0c\u4ee5\u7372\u53d6\u66f4\u597d\u7684\u8a13\u7df4\u8cc7\u6599\u3002\u5f9e\u5be6\u9a57\u7d50\u679c\u4e2d\u53ef\u4ee5\u770b\u51fa\uff0c \u7d93\u904e\u7be9\u9078\u5f8c\u8a13\u7df4\u7684\u6a21\u578b\u6548\u80fd\u4f86\u5230 0.788\uff0c\u5f9e 0.665 \u5927\u5e45\u63d0\u5347 15.6%\u3002\u800c\u5229\u7528\u4eba\u540d\u8fa8\u8b58\u6a21\u578b \u53bb\u9664\u8a3b\u518a\u4eba\u540d\u7684\u65b9\u6cd5\u6bd4\u6b63\u898f\u8868\u9054\u5f0f\u4f86\u7684\u66f4\u597d\u3002\u6b64\u5916\uff0c\u53bb\u9664\u4e0d\u5305\u542b\u8208\u8da3\u9ede\u7684\u65b9\u6cd5\u53ef\u4ee5\u7bc0\u7701 \u6211\u5011\u5c07 L \u548c \u5011\u5c07\u8208\u8da3\u9ede\u4f9d\u9577\u5ea6\u7531\u5927\u81f3\u5c0f\u6392\u5217\uff0c\u4e26\u5c07\u6bd4\u5c0d\u6210\u529f\u7684\u90e8\u4efd\u53bb\u9664\u5f8c\uff0c\u518d\u6bd4\u5c0d\u8f03\u77ed\u7684\u5be6\u9ad4\u3002\u8209 \u4f8b\u800c\u8a00\uff0c \u300c\u570b\u7acb\u4e2d\u592e\u5927\u5b78\u7684\u5c0f\u8c6c\u6ef7\u5473\u5f88\u597d\u5403\u300d\u6bd4\u5c0d\u5230\u300c\u570b\u7acb\u4e2d\u592e\u5927\u5b78\u300d\u4e4b\u5f8c\uff0c\u53e5\u5b50\u7e2e\u6e1b \u6210\u300c\u7684\u5c0f\u8c6c\u6ef7\u5473\u5f88\u597d\u5403\u300d \uff0c\u518d\u8207\u5269\u9918\u7684\u8208\u8da3\u9ede\u6bd4\u5c0d\uff0c\u82e5\u662f\u7e2e\u6e1b\u5f8c\u7684\u53e5\u5b50\u9577\u5ea6\u5c0f\u65bc 2\uff0c\u5247 \u53ef\u76f4\u63a5\u7d50\u675f\u6bd4\u5c0d\u6d41\u7a0b\u3002 \u5716\u4e09\u3001\u81ea\u52d5\u6a19\u8a18\u7bc4\u4f8b\u5716(\u25bd\u548c\u25bc\u7b26\u865f\u4ee3\u8868\u5be6\u9ad4\u7684\u958b\u59cb\u4ee5\u53ca\u7d50\u675f) 3.3 \u7279\u5fb5\u64f7\u53d6 2 Before_2 bigram word before entity or not 3 Before_3 trigram word before entity or not 4 Prefix_1 prefix unigram word or not 5 Prefix_2 prefix bigram word or not 6 Prefix_3 B I E O SumL2 \u5be6\u9a57\u7684\u7b2c\u4e8c\u90e8\u4efd\u5247\u662f POI Relation \u7684\u9810\u6e2c\u3002\u6211\u5011\u5f9e\u9ec3\u9801\u4e0a\u641c\u96c6\u4e26\u7be9\u9078 4,000 \u51fa\u500b\u6b63\u78ba\u7684 \u7b54\u6848\u5206\u985e\uff0c\u800c\u7b2c\u4e8c\u7a2e\u548c\u7b2c\u4e09\u7a2e\u65b9\u6cd5\u4eff\u6548\u771f\u5be6\u7cfb\u7d71\u904b\u4f5c\u72c0\u6cc1\uff0c\u5305\u542b\u548c\u6b63\u4f8b\u8f03\u76f8\u8fd1\u7684\u53cd\u4f8b\uff0c \u51fa\uff0c\u7576\u95dc\u806f\u5206\u985e\u6a21\u578b\u80fd\u6210\u529f\u5206\u985e\u51fa\u6b63\u78ba\u7684\u914d\u5c0d\u6642\uff0c\u642d\u914d\u4e0a\u8d8a\u597d\u7684\u8208\u8da3\u9ede\u8fa8\u8b58\u6a21\u578b\u80fd\u5920\u63d0 \u8207\u8208\u8da3\u9ede\u4f5c\u70ba\u95dc\u9375\u5b57\u7684\u641c\u5c0b\u7d50\u679c\u6578\u8d8a\u591a\uff0c\u4ee3\u8868\u5730\u5740\u8207\u8208\u8da3\u9ede\u9593\u7684\u95dc\u806f\u6027\u8d8a\u9ad8\u3002\u7279\u5fb5\u56db\u548c \u7279\u5fb5\u4e94\u900f\u904e\u8a08\u7b97\u689d\u4ef6\u6a5f\u7387\u53d6\u5f97\uff0c\u548c\u7279\u5fb5\u4e09\u6709\u8457\u76f8\u540c\u6027\u8cea\uff0c\u6578\u503c\u8d8a\u5927\u4ee3\u8868\u95dc\u806f\u6027\u8d8a\u5927\u3002 Labler2 B 5,161 28 0 621 5,810 I 60 24,483 232 3,204 \u5730\u5740\u8207\u8208\u8da3\u9ede\u914d\u5c0d\u505a\u70ba\u6b63\u4f8b(Positive example)\uff0c\u800c\u53cd\u4f8b(Negative example)\u7684\u6311\u9078\u5206\u6210\u4e09 \u8209\u4f8b\u4f86\u8aaa\u300c\u5168\u5bb6(\u4e2d\u592e\u5e97)\u300d\u662f\u6b63\u78ba\u7b54\u6848\uff0c\u800c\u300c\u5168\u5bb6\u300d\u51fa\u73fe\u5728\u53cd\u4f8b\u4e2d\uff0c\u5373\u6709\u53ef\u80fd\u5f71\u97ff\u7d50\u679c\u3002 \u9ad8\u6548\u80fd\uff0c\u6e1b\u5c11\u8fa8\u8b58\u932f\u8aa4\u8208\u8da3\u9ede\u7684\u6a5f\u7387\u4e26\u5927\u5e45\u964d\u4f4e\u5be6\u9a57\u6642\u9593\uff0c\u540c\u6642\u4e5f\u63d0\u5347\u6574\u9ad4\u7cfb\u7d71\u6548\u7387\u3002 27,979 E 0 80 4,937 793 5,810 \u7a2e\uff1a\u5f9e\u6b63\u4f8b\u4e2d\u96a8\u6a5f\u6311\u9078\u8208\u8da3\u9ede\u53bb\u548c\u5730\u5740\u505a\u914d\u5c0d\uff0c\u82e5\u914d\u5c0d\u7d50\u679c\u70ba\u6b63\u4f8b\u5247\u53bb\u9664\u4e26\u91cd\u65b0\u96a8\u6a5f\u6311 \u8868\u56db\u3001\u7cfb\u7d71\u6e2c\u8a66\u7d50\u679c\u6bd4\u8f03\u8868(Efficiency for correct POI = Accuracy * Efficiency) \u6700\u5f8c\u6211\u5011\u7684\u7cfb\u7d71\u80fd\u5f9e\u641c\u5c0b\u7d50\u679c\u4e2d\u8fa8\u8b58\u5730\u5740\uff0c\u900f\u904e\u8fa8\u8b58\u51fa\u8a72\u5730\u5740\u9644\u8fd1\u7684\u8208\u8da3\u9ede\uff0c\u518d\u5229\u7528\u95dc prefix trigram word or not 7 Sufffix_1 suffix unigram word or not 8 Sufffix_2 suffix bigram word or not 9 Sufffix_3 suffix trigram word or not 10 After_1 unigram word after entity or not 11 After_2 bigram word after entity or not 12 After_3 trigram word after entity or not 13 English/Number English or number? 14 Symbol Symbol or not? \u56db\u3001\u5730\u5740\u8207\u8208\u8da3\u9ede\u95dc\u806f\u5206\u985e \u7279\u5fb5\u516d\u5230\u7279\u5fb5\u516b\u63a1\u7528 co-occurrence \u7684\u65b9\u6cd5\u8a08\u7b97\uff0c\u5229\u7528\u5730\u5740\u3001\u8208\u8da3\u9ede\u6216\u662f\u5730\u5740\u8207\u8208\u8da3\u9ede \u4f5c\u70ba\u95dc\u9375\u5b57\u7684\u641c\u5c0b\u7d50\u679c\uff0c\u8a08\u7b97\u5730\u5740\u548c\u8208\u8da3\u9ede\u540c\u6642\u51fa\u73fe\u7684\u6a5f\u7387\uff0c\u8a72\u7279\u5fb5\u503c\u8d8a\u5927\u4ee3\u8868\u5169\u8005\u4e00 \u8d77\u63d0\u5230\u7684\u6a5f\u7387\u8d8a\u5927\uff0c\u95dc\u806f\u6027\u4e5f\u5c31\u6108\u9ad8\u3002 \u82e5\u8a72\u5730\u5740\u51fa\u73fe\u5728\u8a72\u8208\u8da3\u9ede\u7684\u641c\u5c0b\u7d50\u679c\u4e4b\u7b2c\u4e00\u7bc7\uff0c\u4ee3\u8868\u5169\u8005\u7684\u95dc\u806f\u8d8a\u9ad8\uff1b\u53cd\u4e4b\u82e5\u5728\u6700\u5f8c\u4e00 \u7bc7\u624d\u6709\u63d0\u53ca\u3001\u751a\u81f3\u6c92\u6709\u51fa\u73fe\uff0c\u5247\u4ee3\u8868\u95dc\u806f\u8f03\u4f4e\u3002\u56e0\u6b64\u6211\u5011\u63a1\u7528 NDCG \u4f5c\u70ba\u7b2c\u4e5d\u548c\u7b2c\u5341 \u500b\u7279\u5fb5\u3002NDCG \u662f\u7a2e\u8a08\u7b97\u6392\u540d\u7684\u65b9\u6cd5\uff0c\u5e38\u7528\u4f86\u6e2c\u91cf\u641c\u5c0b\u5f15\u64ce\u7684\u6f14\u7b97\u6cd5\u662f\u5426\u6709\u6548\u3002 \u7b2c\u5341\u4e00\u500b\u7279\u5fb5\u662f\u9918\u5f26\u76f8\u4f3c\u5ea6\uff0c\u6578\u503c\u8d8a\u9ad8\u4ee3\u8868\u5169\u8005\u7684\u76f8\u4f3c\u5ea6\u8d8a\u9ad8\uff0c\u95dc\u806f\u6027\u4e5f\u5c31\u8d8a\u9ad8\u3002\u6700\u5f8c O 583 2,025 635 367,859 371,102 SumL1 5,804 26,616 5,804 372,477 410,701 \u8208\u8da3\u9ede\u8fa8\u8b58\u7684\u6e2c\u8a66\u8cc7\u6599\u63a1\u7528\u4eba\u5de5\u6a19\u8a18\uff0c\u4ee5 250 \u500b\u985e\u5225\u540d\u7a31\u30011,000 \u500b\u98df\u7269\u540d\u7a31\u30011,000 \u5716\u516d\u3001LSP1 \u6a21\u578b\u4e4b\u4e0d\u540c\u9577\u5ea6\u5be6\u9ad4\u6548\u80fd\u5716 5.1.2 \u77ed\u8208\u8da3\u9ede\u6548\u80fd\u63d0\u5347 \u5716\u4e5d\u3001\u589e\u52a0\u77ed\u5be6\u9ad4\u8a13\u7df4\u8cc7\u6599\u91cf\u5c0d\u9577\u5ea6\u70ba 3 \u8208\u8da3\u9ede\u4e4b \u5716\u5341\u3001\u589e\u52a0\u77ed\u5be6\u9ad4\u8a13\u7df4\u8cc7\u6599\u91cf\u5c0d\u9577\u5ea6\u70ba 4 \u8208\u8da3\u9ede\u4e4b \u5b78\u7fd2\u66f2\u7dda\u5716 # of POI # of candidate Accuracy Cost (min) Efficiency Efficiency for correct POI \u806f\u5206\u985e\u6a21\u578b\u914d\u5c0d\u5730\u5740\u8207\u8208\u8da3\u9ede\uff0c\u627e\u5230\u8a72\u5730\u5740\u6700\u6709\u53ef\u80fd\u7684\u8208\u8da3\u9ede\uff0c\u9054\u5230\u81ea\u52d5\u64f4\u5145\u4ee5\u53ca\u81ea\u52d5 \u9078\uff0c\u76f4\u5230\u53cd\u4f8b\u6578\u91cf\u9054\u5230 4,000 \u500b\u3002\u7b2c\u4e8c\u7a2e\u5247\u662f\u6a21\u4eff\u7cfb\u7d71\u771f\u5be6\u904b\u4f5c\u6642\u5f9e\u5730\u5740\u7684\u641c\u5c0b\u7d50\u679c\u4e2d (POI/day) (POI/day) \u6316\u6398\u7684\u529f\u80fd\u3002\u5176\u6548\u7387\u9054\u5230\u6bcf\u500b IP \u6bcf\u5929\u80fd\u722c\u53d6\u7d04 49 \u500b\u65b0\u7684\u8208\u8da3\u9ede\u3002 \u8fa8\u8b58\u5230\u5176\u4ed6\u8208\u8da3\u9ede\u4e26\u8207\u5176\u5730\u5740\u914d\u5c0d\u6210\u7684\u53cd\u4f8b\uff1b\u6700\u5f8c\u5247\u662f\u7d9c\u5408\u5169\u7a2e\u65b9\u6cd5\u505a\u70ba\u7b2c\u4e09\u7a2e\u4e0d\u540c\u7684 \u5b78\u7fd2\u66f2\u7dda\u5716 ALL + Random 816 277 0.291 2,307 53.70 15.63 \u5728\u672a\u4f86\u7814\u7a76\u4e0a\uff0c\u9996\u5148\u6211\u5011\u7684\u8208\u8da3\u9ede\u5be6\u9ad4\u7be9\u9078\u53ea\u4fdd\u7559\u62ec\u865f\uff0c\u5176\u9918\u7b26\u865f\u4ee5\u53ca\u82f1\u6587\u6578\u5b57\u7686\u88ab\u53bb \u6e96\u5099\u65b9\u5f0f\u3002\u6e2c\u8a66\u8cc7\u6599\u5247\u6311\u9078\u548c\u8a13\u7df4\u8cc7\u6599\u4e0d\u540c\u7684 2,500 \u500b\u6b63\u4f8b\uff0c\u4f7f\u7528\u7b2c\u4e00\u7a2e\u548c\u7b2c\u4e8c\u7a2e\u65b9\u5f0f \u500b\u5730\u5740\u4ee5\u53ca 1,000 \u500b\u8208\u8da3\u9ede\u505a\u70ba\u641c\u5c0b\u95dc\u9375\u5b57\uff0c\u5171\u722c\u53d6 4,000 \u500b\u641c\u5c0b\u7d50\u679c\u3002\u5148\u4ee5\u81ea\u52d5\u6a19\u8a18 \u7684\u65b9\u6cd5\u5c0d\u8a13\u7df4\u8cc7\u6599\u9032\u884c\u7b54\u6848\u6a19\u8a18\uff0c\u518d\u8acb\u5169\u4f4d\u6a19\u8a18\u4eba\u54e1\u4fee\u6b63\u932f\u8aa4\u4ee5\u53ca\u88dc\u6a19\u7b54\u6848\u3002\u6a19\u8a18\u4e4b\u4e00 \u81f4\u6027\u4fe1\u5ea6(Kappa)\u503c\u70ba 0.886\uff0c\u8868\u793a\u6a19\u8a18\u7b54\u6848\u7684\u53ef\u4fe1\u5ea6\uff0c\u5982\u8868\u4e09\u932f\u8aa4! \u627e\u4e0d\u5230\u53c3\u7167\u4f86\u6e90\u3002 \u6240\u793a\u3002 5.1.1 \u8208\u8da3\u9ede\u7be9\u9078\u6548\u80fd\u5206\u6790 \u5f9e\u6574\u9ad4\u6548\u80fd\u7684\u89d2\u5ea6\u4f86\u770b\uff0c\u589e\u52a0\u77ed\u8208\u8da3\u9ede\u8cc7\u6599\u91cf\u53ef\u4f7f\u7cbe\u6e96\u7387\u5f9e 0.799 \u63d0\u5347\u81f3 0.855(\u5982\u5716\u5341 \u5404\u6e96\u5099 1,250 \u500b\u53cd\u4f8b\u4e26\u9032\u884c\u4eba\u5de5\u6a19\u8a18\uff0c\u6700\u7d42\u8cc7\u6599\u5171\u5305\u542b 2,740 \u500b\u6b63\u4f8b\u4ee5\u53ca 2,560 \u500b\u53cd\u4f8b\u3002 ALL + Special 816 13 0.182 2,311 1.25 0.23 \u9664\uff0c\u7136\u800c\u5be6\u969b\u4e0a\u6709\u8a31\u591a\u8208\u8da3\u9ede\u662f\u4e2d\u82f1\u6df7\u96dc\u6216\u662f\u593e\u5e36\u6578\u5b57\uff0c\u672a\u4f86\u82e5\u80fd\u8a13\u7df4\u51fa\u8de8\u8a9e\u8a00\u7684\u8fa8\u8b58 \u70ba\u6539\u5584\u8208\u8da3\u9ede\u8fa8\u8b58\u6a21\u578b\u6548\u80fd\uff0c\u6211\u5011\u5206\u5225\u91dd\u5c0d\u4e0d\u540c\u9577\u5ea6\u7684\u5be6\u9ad4\u9032\u884c\u5206\u6790\uff0c\u4ee5\u5148\u524d\u63d0\u53ca\u7684\u6e05 \u4e00)\uff1b\u53ec\u56de\u7387\u96d6\u6c92\u6709\u660e\u986f\u6210\u9577\uff0c\u4e5f\u5f9e 0.777 \u4f86\u5230 0.781\uff1bF1 \u503c\u5247\u662f\u5f9e 0.788 \u6539\u5584\u81f3 0.816\u3002 ALL + Mix 816 3 1 2,306 1.25 1.25 \u6a21\u578b\uff0c\u5c31\u80fd\u8fa8\u8b58\u51fa\u66f4\u591a\u8208\u8da3\u9ede\u3002\u5176\u6b21\u96d6\u7136\u4e09\u7a2e\u4e0d\u540c\u7684\u95dc\u806f\u5206\u985e\u6a21\u578b\u6548\u80fd\u5dee\u7570\u4e26\u4e0d\u5927\uff0c\u4f46 \u55ae L \u548c\u6e05\u55ae SP \u722c\u53d6\u7b2c\u4e00\u7b46\u641c\u5c0b\u7d50\u679c\u547d\u540d\u70ba LSP1\uff0c\u5982\u5716\u516d\u6240\u793a\u3002\u53ef\u4ee5\u770b\u5230\u9577\u5ea6\u4e09\u548c\u56db \u5f9e\u5716\u5341\u4e09\u4e2d\u53ef\u4ee5\u770b\u51fa\u4e09\u500b\u6a21\u578b\u7684\u6548\u80fd\u5dee\u4e0d\u591a\uff0c\u5176\u4e2d\u7b2c\u4e8c\u7a2e\u7684\u53cd\u4f8b\u6e96\u5099\u65b9\u5f0f\u6548\u80fd\u6700\u4f4e\uff0c\u6e96 \u5728\u7cfb\u7d71\u6e2c\u8a66\u6642\u537b\u6709\u975e\u5e38\u5927\u7684\u4e0d\u540c\uff0c\u5f9e\u5b78\u7fd2\u66f2\u7dda\u4e2d\u53ef\u4ee5\u770b\u51fa\uff0c\u8a13\u7df4\u8cc7\u6599\u4e94\u767e\u6642\u6e96\u78ba\u7387\u5373\u9054 \u7684\u77ed\u5be6\u9ad4\u7684\u6548\u80fd\u4f4e\u65bc\u6574\u9ad4\u6548\u80fd\uff0c\u56e0\u6b64\u6211\u5011\u5617\u8a66\u5169\u7a2e\u65b9\u6cd5\u53bb\u63d0\u5347\u77ed\u5be6\u9ad4\u8fa8\u8b58\u6548\u80fd\uff1a\u5408\u4f75\u6a21 \u78ba\u7387\u70ba 0.744\uff0c\u4f46\u662f\u6df7\u5408\u8a13\u7df4\u8cc7\u6599\u7684\u6548\u679c\u5247\u5f97\u5230\u6700\u9ad8\u7684 0.754 \u6e96\u78ba\u7387\u3002\u554f\u984c\u53ef\u80fd\u51fa\u5728\u8a13 LSP10 + Random 340 88 0.648 1,034 75.20 48.73 \u5230 0.739\uff0c\u82e5\u80fd\u624b\u52d5\u6a19\u8a18\u5f9e\u5730\u5740\u7684\u641c\u5c0b\u7d50\u679c\u4e2d\u8fa8\u8b58\u975e\u6b63\u4f8b\u914d\u5c0d\u7684\u8208\u8da3\u9ede\uff0c\u5c07\u539f\u672c\u88ab\u8996\u70ba \u578b\u4ee5\u53ca\u589e\u52a0\u77ed\u5be6\u9ad4\u8cc7\u6599\u91cf\u3002 LSP10 + Special 340 2 1 1,033 1.39 1.39 \u53cd\u4f8b\u7684\u8208\u8da3\u9ede\u4fee\u6b63\u6210\u6b63\u4f8b(\u4f8b\u5982\u516c\u53f8\u7e2e\u5beb\u6216\u662f\u9023\u9396\u5e97)\uff0c\u6216\u8a31\u80fd\u63d0\u5347\u6a21\u578b\u6548\u80fd\uff0c\u5728\u7cfb\u7d71\u9762 \u7df4\u8cc7\u6599\u7684\u6e96\u5099\u4e2d\uff0c\u8fa8\u8b58\u51fa\u7684\u8208\u8da3\u9ede\u6216\u8a31\u7684\u78ba\u843d\u5728\u8a72\u5730\u5740\u4e0a\uff0c\u7136\u800c\u8a72\u914d\u5c0d\u4e26\u4e0d\u5b58\u5728\u9ec3\u9801\u4e2d \u8cc7\u6599\u4f86\u6e90\u4e94\u82b1\u516b\u9580\uff0c\u53ef\u80fd\u4f86\u81ea\u65b0\u805e\u3001\u90e8\u843d\u683c\u6587\u7ae0\u6216\u662f\u793e\u7fa4\u7684\u8cbc\u6587\uff0c\u6bcf\u500b\u53e5\u5b50\u95e1\u8ff0\u7684\u5167\u5bb9 \u6b64\u8655\u4f7f\u7528\u9017\u865f\u4ee5\u53ca\u53e5\u865f\u505a\u70ba\u65b7\u53e5\u7684\u4f9d\u64da\uff0c\u5728\u8cc7\u6599\u91cf\u6700\u5927\u7684\u5be6\u9a57\u4e2d\uff0c\u65b7\u53e5\u5f8c\u6709 5,323,009 \u53e5\uff0c\u7e3d\u5b57\u6578\u66f4\u662f\u7834\u5104\u5b57\uff0c\u57fa\u65bc\u8a2d\u5099\u4ee5\u53ca\u8a13\u7df4\u901f\u5ea6\u7684\u8003\u91cf\uff0c\u6211\u5011\u53ea\u4fdd\u7559\u542b\u6709\u8208\u8da3\u9ede\u7684\u53e5\u5b50\uff0c \u4e5f\u5c31\u662f\u53bb\u9664\u6240\u6709\u7684\u4e0d\u5305\u542b\u8208\u8da3\u9ede\u53e5\u5b50\u7684\u8ca0\u7bc4\u4f8b(negative example)\u3002\u7d93\u904e\u5be6\u9a57\u5f8c\u767c\u73fe\u6548\u80fd \u672c\u7bc0\u7684\u76ee\u7684\u5373\u5728\u5224\u65b7\u7d66\u5b9a\u5730\u5740 a \u8207\u8208\u8da3\u9ede p \u4e4b\u9593\u7684\u914d\u5c0d\u95dc\u4fc2\u3002\u57fa\u672c\u4e0a\u7cfb\u7d71\u5c07\u5c0d\u641c\u5c0b\u5f15\u64ce \u5206\u5225\u9001\u51fa\u4e09\u500b\u67e5\u8a62\uff1aa\u3001p\u3001\u53ca a+p \u4ee5\u5f97\u5230\u641c\u5c0b\u7d50\u679c\uff1aT a \u4ee3\u8868\u5730\u5740\u7684\u524d\u5341\u7bc7\u641c\u5c0b\u7d50\u679c\uff0c T p \u4ee3\u8868\u8208\u8da3\u9ede\u7684\u524d\u5341\u7bc7\u641c\u5c0b\u7d50\u679c\uff0cT a+p \u4ee3\u8868\u5730\u5740\u8207\u8208\u8da3\u9ede\u7684\u524d\u5341\u7bc7\u641c\u5c0b\u7d50\u679c\u3002\u70ba\u4e86\u6709\u6548 \u793a\u8cc7\u6599\u8d8a\u65b0\uff0c\u6b63\u78ba\u6027\u4ea6\u6703\u8f03\u9ad8\u3002 \u4e94\u3001\u5be6\u9a57 \u4f5c\u70ba\u5df2\u77e5\u8208\u8da3\u9ede\u5229\u7528\u81ea\u52d5\u6a19\u8a18\u6240\u5f97\u7684\u8fa8\u8b58\u6548\u80fd(\u5716\u56db) \uff0c\u4e26\u8207\u81ea\u52d5\u6a19\u8a18\u6240\u8a13\u7df4\u51fa\u7684\u6a21\u578b (\u5716\u4e94\u3001\u4e09\u7a2e\u5be6\u9ad4\u6a21\u578b\u8fa8\u8b58\u6548\u80fd\u6bd4\u8f03\u5716)\u505a\u6bd4\u8f03\u3002\u5f9e\u5716\u56db\u4e2d\u53ef\u4ee5\u770b\u51fa\uff0c\u50c5\u7528\u5b57\u5178\u6bd4\u5c0d\u65b9 \u6cd5\u7684\u6548\u80fd\u6709\u9650\uff0c\u6c92\u6709\u7be9\u9078\u904e\u7684\u539f\u59cb\u9ec3\u9801\u5546\u5bb6\u50c5\u6709 0.291 \u7684\u6e96\u78ba\u7387\uff0c\u96d6\u7136\u7be9\u9078\u904e\u4eba\u540d\u7684\u6e96 \u78ba\u7387\u53ef\u9054 0.785 \u53ca 0.778\uff0c\u4f46\u662f\u53ec\u56de\u7387\u4e0d\u5230 0.3\uff1b\u5176\u4e2d\u4f7f\u7528 LSR \u6e05\u55ae\u7684\u6700\u4f73\u6548\u80fd\uff0cF1 \u503c \u56db\u5be6\u9ad4\u4ee5\u53ca\u6574\u9ad4\u7684\u7cbe\u6e96\u7387\u7686\u4f4e\u65bc\u4e00\u822c\u6a21\u578b(LSP1)\uff0c\u5373\u4f7f\u53ec\u56de\u7387\u90fd\u6bd4\u4e00\u822c\u6a21\u578b\u4f86\u7684\u597d\uff0c\u6574 \u9ad4\u770b\u4f86 F1 \u503c\u90fd\u8f03\u4f4e\uff0c\u56e0\u6b64\u5408\u4f75\u6a21\u578b\u7684\u65b9\u5f0f\u7121\u6cd5\u6709\u6548\u6539\u5584\u6548\u80fd\u3002 \u81ea\u52d5\u6a19\u8a18\u4f5c\u70ba\u57fa\u790e\uff0c\u6e96\u5099\u8a13\u7df4\u8cc7\u6599\uff0c\u518d\u5229\u7528 CRF++\u8a13\u7df4\u7684\u8fa8\u8b58\u6a21\u578b\u3002\u7531\u65bc\u4e2d\u83ef\u9ec3\u9801\u4e0a \u6e2c\u8a66\u3002\u7531\u65bc\u4e00\u500b\u5730\u5740\u9644\u8fd1\u53ef\u80fd\u6703\u6709\u63d0\u5230\u6578\u500b\u8208\u8da3\u9ede(\u8868\u56db\u7b2c 2 \u884c)\uff0c\u6211\u5011\u4fdd\u7559\u9a57\u8b49\u6a5f\u7387\u5927 5.1.3 \u8a13\u7df4\u6a21\u578b\u6548\u7387\u4e4b\u63d0\u5347 \u6211\u5011\u4ee5\u9ec3\u9801\u5546\u5bb6\u8208\u8da3\u9ede\u505a\u70ba\u5df2\u77e5\u5be6\u9ad4\u540d\u7a31\uff0c\u61c9\u7528\u641c\u5c0b\u5f15\u64ce\u6536\u96c6\u5305\u542b\u8208\u8da3\u9ede\u7684\u53e5\u5b50\uff0c\u4e26\u4ee5 \u6700\u5f8c\u6211\u5011\u5229\u7528\u5169\u500b\u8208\u8da3\u9ede\u8fa8\u8b58\u6a21\u578b(ALL, LSP10)\u4ee5\u53ca\u4e09\u500b\u95dc\u806f\u6a21\u578b\u9032\u884c\u5171\u516d\u6b21\u7684\u7cfb\u7d71 \u5716\u5341\u4e00\u3001\u589e\u52a0\u77ed\u8208\u8da3\u9ede\u8a13\u7df4\u8cc7\u6599\u4e4b\u6574\u9ad4\u6548\u80fd\u6bd4\u8f03\u5716 \u516d\u3001\u7d50\u8ad6 \u53e5\uff1bSP \u6709 65,372 \u500b\u5be6\u9ad4\u3001108,874 \u53e5\u3002\u5f9e\u5716\u4e03\u53ef\u4ee5\u770b\u51fa\uff0c\u5408\u4f75\u6a21\u578b(E_LSP1)\u5728\u9577\u5ea6\u4e09\u3001 5.3 \u7cfb\u7d71\u6548\u80fd\u6e2c\u8a66 \u4e2d\u8fa8\u8b58\u6642\u9593\uff0c\u6e1b\u7cfb\u7d71\u6642\u9593\u5f8c\u53d6\u5c0d\u6578\u5f97\u5230\u6b64\u7279\u5fb5\u503c\uff0c\u8a72\u503c\u8d8a\u5927\u8868\u793a\u8d8a\u820a\uff1b\u53cd\u4e4b\u8a72\u503c\u8d8a\u5c0f\u8868 \u5f9e\u64f7\u53d6\u51fa\u65b0\u7684\u5730\u5740\u7db2\u9801\u4e2d\u8fa8\u8b58\u53ef\u80fd\u7684\u8208\u8da3\u9ede\u540d\u7a31\uff0c\u6700\u5f8c\u9032\u884c\u914d\u5c0d\u95dc\u4fc2\u9810\u6e2c (\u5982\u5716\u4e00\u6240\u793a) \u3002 \u518d\u806f\u96c6\u5169\u500b\u6a21\u578b\u7684\u7b54\u6848\uff0c\u505a\u70ba\u6700\u7d42\u7684\u6a19\u8a18\u7d50\u679c\u3002\u5176\u4e2d L \u5171\u6709 491,330 \u500b\u5be6\u9ad4\u3001773,927 \u4e0d\u76e1\u76f8\u540c\uff0c\u56e0\u6b64\u6211\u5011\u5229\u7528\u8208\u8da3\u9ede\u524d\u5f8c\u7684\u5b57\u8a5e\u505a\u70ba\u7279\u5fb5\uff0c\u518d\u7528 CRF++\u9032\u884c\u6a21\u578b\u7684\u8a13\u7df4\u3002 \u5982\u7dd2\u8ad6\u6240\u8ff0\uff0c\u672c\u8ad6\u6587\u9996\u5148\u61c9\u7528\u5730\u5740\u95dc\u9375\u5b57\u5c0d\u641c\u5c0b\u5f15\u64ce\u67e5\u8a62\u5f97\u5230\u53ef\u80fd\u5305\u542b\u5730\u5740\u7684\u7db2\u9801\uff0c\u518d \u4e00\u500b\u7279\u5fb5\u548c\u6642\u9593\u6709\u95dc\uff0c\u6211\u5011\u8a8d\u70ba\u8d8a\u65b0\u7684\u8cc7\u8a0a\u8d8a\u6b63\u78ba\u3002\u56e0\u6b64\uff0c\u5229\u7528\u6b63\u898f\u8868\u9054\u5f0f\u5f9e\u641c\u5c0b\u7d50\u679c \u6211\u5011\u5229\u7528\u5148\u524d\u63d0\u53ca\u7684\u6e05\u55ae L \u548c\u6e05\u55ae SP \u722c\u53d6\u7b2c\u4e00\u7b46\u641c\u5c0b\u7d50\u679c(LSP1)\uff0c\u4e26\u5206\u5225\u8a13\u7df4\u6a21\u578b\uff0c \u7684\u8cc7\u6599\u4e2d\uff0c\u56e0\u6b64\u88ab\u6211\u5011\u505a\u70ba\u53cd\u4f8b\u9032\u800c\u964d\u4f4e\u6548\u80fd\u3002 LSP10 + Mix 340 2 1 1,034 1.39 1.39 \u6642\u4ea6\u80fd\u66f4\u7b26\u5408\u5be6\u969b\u72c0\u6cc1\uff0c\u627e\u5230\u66f4\u591a\u66f4\u6b63\u78ba\u7684\u5019\u9078\u4eba\u3002\u6b64\u5916\uff0c\u5229\u7528 Google \u641c\u5c0b\u5f15\u64ce\u6703\u6709
\u4e0d\u540c\u7684\u5730\u5716\u670d\u52d9\u53ef\u80fd\u6709\u5404\u81ea\u7684\u7279\u8272\uff0c\u4f46\u591a\u6578\u90fd\u6703\u63d0\u4f9b\u8208\u8da3\u9ede\u7684\u5730\u5740\uff0c\u82e5\u8208\u8da3\u9ede\u70ba\u5e97\u5bb6\u6216 \u662f\u6a5f\u69cb\u5247\u591a\u6703\u589e\u52a0\u96fb\u8a71\u3001\u71df\u696d\u6642\u9593\u7b49\u8cc7\u8a0a\u3002\u800c\u5730\u5716\u670d\u52d9\u8457\u91cd\u5728\u8cc7\u8a0a\u7684\u6b63\u78ba\u6027\u4e0a\uff0c\u82e5\u63d0\u4f9b \u7684\u8cc7\u8a0a\u8207\u4e8b\u5be6\u4e0d\u7b26\uff0c\u5c07\u6703\u9020\u6210\u4f7f\u7528\u8005\u56f0\u64fe\uff0c\u5bb9\u6613\u56e0\u70ba\u932f\u8aa4\u8cc7\u8a0a\u800c\u88ab\u7d66\u4e88\u8f03\u4f4e\u5206\u7684\u8a55\u50f9\uff0c \u5c0e\u81f4\u53e3\u7891\u4e0d\u4f73\u7121\u4eba\u4f7f\u7528\u7684\u7a98\u5883\u3002 \u5730\u5716\u670d\u52d9\u7684\u554f\u984c\u662f\u5982\u4f55\u8490\u96c6\u65b0\u7684\u8cc7\u6599\u3002\u6839\u64da\u8ca1\u653f\u90e8\u7684\u7d71\u8a08\uff0c\u6211\u570b\u9910\u98f2\u696d\u4e4b\u71df\u5229\u4e8b\u696d\u5bb6\u6578 81.6%\u7684 F1 \u503c\uff0c\u53e6\u5916\u5730\u5740\u8207\u5546\u5bb6\u914d\u5c0d\u9a57\u8b49\u6a21\u578b\u7cbe\u6e96\u7387\u70ba 74.56%\uff0c\u7cfb\u7d71\u6bcf\u5929\u80fd\u81ea\u52d5\u5316\u627e \u5230 49 \u500b\u65b0\u7684\u8208\u8da3\u9ede\u3002 \u672c\u8ad6\u6587\u7684\u5167\u5bb9\u7d44\u7e54\u5982\u4e0b\uff1a\u7b2c\u4e8c\u7ae0\u63cf\u8ff0\u76f8\u95dc\u7814\u7a76\u3002\u7b2c\u4e09\u7ae0\u70ba\u7cfb\u7d71\u67b6\u69cb\u53ca POI \u540d\u7a31\u8fa8\u8b58\uff0c \u7b2c\u56db\u7ae0\u70ba POI \u8207\u5730\u5740\u95dc\u4fc2\u9810\u6e2c\u6a21\u578b\uff0c\u7b2c\u4e94\u7ae0\u70ba\u5be6\u9a57\u6578\u64da\u4ee5\u53ca\u5be6\u9a57\u7d50\u679c\uff0c\u6700\u5f8c\u7b2c\u516d\u7ae0\u63d0 \u9664\u4e86\u7b2c\u5341\u4e09\u548c\u7b2c\u5341\u56db\u500b\u7279\u5fb5\u6709\u56fa\u5b9a\u7684\u5b57\u5178\uff0c\u5269\u9918\u7684\u5b57\u5178\u6703\u5f9e\u8a13\u7df4\u8cc7\u6599\u4e2d\u64f7\u53d6\uff0c\u4e26\u900f\u904e\u8a72 \u51fa\u300c\u7d00\u5ff5\u9928\u300d \uff0c\u6703\u5f97\u5230 0.5 \u5206\uff0c\u90e8\u5206\u6bd4\u5c0d\u5206\u6578\u3001\u7cbe\u6e96\u7387(Precision)\u3001\u53ec\u56de\u7387(Recall)\u4ee5\u53ca \u5716\u4e03\u3001\u5408\u4f75\u6a21\u578b\u77ed\u5be6\u9ad4\u90e8\u5206\u4e4b\u6548\u80fd\u6bd4\u8f03\u5716 \u5716\u516b\u3001\u5408\u4f75\u6a21\u578b\u6574\u9ad4\u4e4b\u6548\u80fd\u6bd4\u8f03\u5716 \u7e3d\u82b1\u8cbb\u6642\u9593\u70ba 1,034 \u5206\u9418\uff0c\u9810\u4f30\u6bcf\u500b IP \u6bcf\u5929\u80fd\u627e\u5230\u7d04 49 \u500b\u65b0\u7684\u8208\u8da3\u9ede\u3002\u800c ALL \u6a21\u578b\u642d \u7c64\u4e4b\u904e\u77ed\u7684\u6b63\u4f8b\u4e5f\u80fd\u63d0\u5347\u6548\u80fd\u3002\u96a8\u8457\u8cc7\u6599\u91cf\u7684\u589e\u52a0\uff0c\u77ed\u5be6\u9ad4\u7684\u6548\u80fd\u7372\u5f97\u6539\u5584\uff0c\u6574\u9ad4 F1 \u5b50\u3001\u81ea\u52d5\u6a19\u8a18\u3001\u8cc7\u6599\u524d\u8655\u88e1\u3001\u7279\u5fb5\u64f7\u53d6\u4ee5\u53ca\u4f7f\u7528 CRF++\u9032\u884c\u6a21\u578b\u8a13\u7df4\uff0c\u5982\u5716\u4e8c\u6240\u793a\u3002 \u5b57\u51fa\u73fe\u7684\u983b\u7387\u9032\u884c\u7be9\u9078\u3002 F1 \u503c\u4e4b\u7b97\u6cd5\u5982\u4ee5\u4e0b\u3002 \u914d\u7b2c\u4e00\u7a2e\u95dc\u806f\u5206\u985e\u6a21\u578b\u7684\u6548\u80fd\u53ea\u6709 0.291\uff0c\u5247\u662f\u56e0\u70ba\u6240\u8fa8\u8b58\u51fa\u7684\u8208\u8da3\u9ede\u4e0d\u6b63\u78ba\u5c0e\u81f4\u6548\u80fd \u503c\u4e5f\u5f9e\u4e00\u500b\u641c\u5c0b\u7d50\u679c\u7684 0.788 \u63d0\u5347\u5230\u5341\u500b\u641c\u5c0b\u7d50\u679c\u7684 0.816\uff0c\u56e0\u6b64\u6211\u5011\u76f8\u4fe1\uff0c\u82e5\u6709\u8db3\u5920 \u6211\u5011\u5f9e\u4e2d\u83ef\u9ec3\u9801\u4e0a\u641c\u96c6 677,172 \u500b\u5546\u5bb6\u8208\u8da3\u9ede\u505a\u70ba\u5be6\u9ad4\u6e05\u55ae\uff0c\u7136\u800c\u4e00\u958b\u59cb\u6240\u5f97\u5230\u7684\u6a21\u578b \u7b2c\u4e8c\u7a2e\u505a\u6cd5\u5247\u662f\u5c07\u77ed\u5be6\u9ad4\u7684\u641c\u5c0b\u7d50\u679c\u5f9e\u4e00\u7b46\u589e\u52a0\u81f3\u4e94\u7b46\u548c\u5341\u7b46\u3002\u5716\u4e5d\u4ee5\u53ca\u5716\u5341\u5206\u5225\u986f\u793a \u4ee5\u53ca\u6548\u7387\u964d\u4f4e\u3002 \u7684\u786c\u9ad4\u8a2d\u5099\uff0c\u8db3\u4ee5\u8ca0\u8377\u66f4\u591a\u7684\u8cc7\u6599\uff0c\u5c31\u80fd\u5920\u8a13\u7df4\u51fa\u66f4\u597d\u7684\u6a21\u578b\u3002 \u6700\u5f8c\u6211\u5011\u5229\u7528 CRF++\u9032\u884c\u8a13\u7df4\uff0c\u5c07\u5e8f\u5217\u6a19\u8a18\u6210 BIEO \u7b26\u865f\uff0cB \u4ee3\u8868\u5be6\u9ad4\u7684\u958b\u59cb\uff0cI \u4ee3\u8868 \u6548\u679c\u4e26\u4e0d\u4f73\uff0c\u539f\u56e0\u662f\u9ec3\u9801\u4e2d\u5546\u5bb6\u540d\u7a31\u53ef\u80fd\u4f7f\u7528\u8a3b\u518a\u4eba\u540d\u3001\u98df\u7269\u540d\u7a31\u3001\u985e\u5225\u540d\u7a31(\u571f\u6728\u5de5 \u7a0b)\u6216\u5730\u5340\u540d\u7a31(\u6843\u5712\u5e02\u4e2d\u58e2\u5340\u3001\u9ad8\u6a39\u9109)\u505a\u70ba\u8208\u8da3\u9ede\uff0c\u5c0e\u81f4\u6240\u5f97\u6a21\u578b\u6a19\u8a18\u6e96\u78ba\u7387\u592a\u4f4e\uff0c\u56e0 \u9577\u5ea6 3 \u53ca 4 \u7684\u5be6\u9ad4\u8fa8\u8b58\u6548\u80fd\uff0c\u6211\u5011\u767c\u73fe\u589e\u52a0\u8a13\u7df4\u8cc7\u6599\u91cf\u53ef\u4ee5\u6709\u6548\u63d0\u5347\u6548\u80fd\uff0cF1 \u503c\u5f9e\u539f \u5be6\u9ad4\u7684\u4e2d\u9593\u5b57\u5143\uff0cE \u4ee3\u8868\u5be6\u9ad4\u7684\u7d50\u675f\uff0cO \u5247\u662f\u4ee3\u8868\u4e0d\u5c6c\u65bc\u5be6\u9ad4\u3002 \u5716\u56db\u3001\u5b57\u5178\u5c0e\u5411\u4e4b\u8208\u8da3\u9ede\u8fa8\u8b58\u6548\u80fd \u5148\u7684 0.387 \u63d0\u5347\u81f3 0.467\u3002\u9577\u5ea6\u70ba\u56db\u7684\u8208\u8da3\u9ede\u8fa8\u8b58\u5247\u662f\u5f9e 0.606 \u63d0\u5347\u81f3 0.676\u3002 \u4e0d\u8ad6\u662f\u5229\u7528\u7b2c\u4e8c\u7a2e\u6216\u7b2c\u4e09\u7a2e\u95dc\u806f\u5206\u985e\u6a21\u578b\u7684\u7d44\u5408\u4e4b\u6548\u7387\u90fd\u6975\u5dee\uff0c\u53ef\u80fd\u662f\u56e0\u70ba\u8a13\u7df4\u8cc7\u6599\u4e2d \u5728\u5730\u5740\u8207\u8208\u8da3\u9ede\u95dc\u806f\u9810\u6e2c\u90e8\u5206\uff0c\u6211\u5011\u63d0\u51fa\u4e09\u7a2e\u4e0d\u540c\u6e96\u5099\u53cd\u4f8b\u7684\u65b9\u6cd5\uff1a\u96a8\u6a5f\u5206\u914d\u8208\u8da3\u9ede\u7d66 \u5716\u4e94\u3001\u4e09\u7a2e\u5be6\u9ad4\u6a21\u578b\u8fa8\u8b58\u6548\u80fd\u6bd4\u8f03\u5716 \u51fa\u7d50\u8ad6\u548c\u672a\u4f86\u7814\u7a76\u65b9\u5411\u3002 \u5716\u5341\u4e8c\u3001\u53bb\u9664\u4e0d\u5305\u542b\u8208\u8da3\u9ede\u4e4b\u6548\u80fd\u53ca\u6642\u9593\u6bd4\u8f03\u5716 \u5716\u5341\u4e09\u3001\u5730\u5740\u8207\u8208\u8da3\u9ede\u95dc\u806f\u6a21\u578b\u6548\u80fd\u6bd4\u8f03\u5716 \u6709\u6b63\u78ba\u914d\u5c0d\u88ab\u8996\u70ba\u53cd\u4f8b\uff0c\u5c0e\u81f4\u5be6\u969b\u6e2c\u8a66\u6642\u6b63\u78ba\u914d\u5c0d\u4e0d\u80fd\u88ab\u6210\u529f\u5206\u985e\uff0c\u5f9e\u5019\u9078\u4eba\u7684\u6578\u91cf\u4e2d \u5730\u5740\u3001\u5f9e\u5730\u5740\u7684\u641c\u5c0b\u7d50\u679c\u64f7\u53d6\u4e4b\u932f\u8aa4\u914d\u5c0d\u8208\u8da3\u9ede\u4ee5\u53ca\u6df7\u5408\u524d\u5169\u8005\uff0c\u5176\u7cbe\u6e96\u7387\u5206\u5225\u70ba 0.746\u3001
", "type_str": "table", "num": null } } } }