| { | |
| "paper_id": "O10-2009", | |
| "header": { | |
| "generated_with": "S2ORC 1.0.0", | |
| "date_generated": "2023-01-19T08:06:49.318307Z" | |
| }, | |
| "title": "Histogram Equalization for Statistical Unknown Word Extraction", | |
| "authors": [ | |
| { | |
| "first": "Yi-Cong", | |
| "middle": [], | |
| "last": "\u9673\u5f08\u7481", | |
| "suffix": "", | |
| "affiliation": {}, | |
| "email": "" | |
| }, | |
| { | |
| "first": "\u570b\uf9f7\u53f0\u7063\u79d1\u6280\u5927\u5b78\u8cc7\u8a0a\u7ba1\uf9e4\u5b78\u7cfb", | |
| "middle": [], | |
| "last": "Chen", | |
| "suffix": "", | |
| "affiliation": {}, | |
| "email": "" | |
| }, | |
| { | |
| "first": "Bor-Shen", | |
| "middle": [], | |
| "last": "\uf9f4\u4f2f\u614e", | |
| "suffix": "", | |
| "affiliation": {}, | |
| "email": "" | |
| }, | |
| { | |
| "first": "\u570b\uf9f7\u53f0\u7063\u79d1\u6280\u5927\u5b78\u8cc7\u8a0a\u7ba1\uf9e4\u5b78\u7cfb", | |
| "middle": [], | |
| "last": "Lin", | |
| "suffix": "", | |
| "affiliation": {}, | |
| "email": "" | |
| } | |
| ], | |
| "year": "", | |
| "venue": null, | |
| "identifiers": {}, | |
| "abstract": "With the evolution of human lives and the accelerated spread of information, new things and concepts are generated quickly, and new words emerge every day. It is therefore important for natural language processing systems to identify new words. This paper used the scheme for Chinese word extraction based on machine learning approaches to combining various statistical features. Due to the broad areas for the natural language applications, however, it is quite probable that the mismatch of statistical characteristics between the training and the testing domains occurs, which degrades the performance for word extraction inevitably. This paper proposes the scheme of utilizing the histogram equalization for feature normalization in statistical approaches. Through this scheme, the mismatch of the feature distributions for the training set and the testing set, with different sizes or in different domains, can be compensated. This makes the statistical approaches of unknown word extraction more robust for novel domains. This scheme was tested on the corpora provided by SIGHAN2. The best results, 68.43% and 71.40% of F-Measure for the CKIP corpus and the HKCU corpus respectively, can be achieved with four features with normalization and histogram equalization. When applied to unknown word extraction in an novel domain, it can be found that this scheme is capable of identifying such pronouns as \"Cape No. 7\"(\u6d77\u89d2\u4e03\u865f), \"Financial Tsunami\"(\u91d1\u878d\u6d77 \u562f) and so on, which are not easy to be extracted by those approaches based on semantic characteristics. This scheme appears not good enough for extracting such new terms as the names of humans, places and organizations, in which the semantic structures are prominent. When compared with the results of unknown word extraction for two Chinese word segmentation systems, it can be observed that this scheme exhibits to be complementary with other approaches, and it is promising to combine approaches with different capabilities.", | |
| "pdf_parse": { | |
| "paper_id": "O10-2009", | |
| "_pdf_hash": "", | |
| "abstract": [ | |
| { | |
| "text": "With the evolution of human lives and the accelerated spread of information, new things and concepts are generated quickly, and new words emerge every day. It is therefore important for natural language processing systems to identify new words. This paper used the scheme for Chinese word extraction based on machine learning approaches to combining various statistical features. Due to the broad areas for the natural language applications, however, it is quite probable that the mismatch of statistical characteristics between the training and the testing domains occurs, which degrades the performance for word extraction inevitably. This paper proposes the scheme of utilizing the histogram equalization for feature normalization in statistical approaches. Through this scheme, the mismatch of the feature distributions for the training set and the testing set, with different sizes or in different domains, can be compensated. This makes the statistical approaches of unknown word extraction more robust for novel domains. This scheme was tested on the corpora provided by SIGHAN2. The best results, 68.43% and 71.40% of F-Measure for the CKIP corpus and the HKCU corpus respectively, can be achieved with four features with normalization and histogram equalization. When applied to unknown word extraction in an novel domain, it can be found that this scheme is capable of identifying such pronouns as \"Cape No. 7\"(\u6d77\u89d2\u4e03\u865f), \"Financial Tsunami\"(\u91d1\u878d\u6d77 \u562f) and so on, which are not easy to be extracted by those approaches based on semantic characteristics. This scheme appears not good enough for extracting such new terms as the names of humans, places and organizations, in which the semantic structures are prominent. When compared with the results of unknown word extraction for two Chinese word segmentation systems, it can be observed that this scheme exhibits to be complementary with other approaches, and it is promising to combine approaches with different capabilities.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Abstract", | |
| "sec_num": null | |
| } | |
| ], | |
| "body_text": [ | |
| { | |
| "text": "T i : \u7b2c i \u500b\u5019\u9078\u8a5e C(T i ) : \u5019\u9078\u8a5e T i \u5728\u6240\u6709\u6587\u4ef6\u4e2d\u51fa\u73fe\u7684\u7e3d\u6b21\uf969 \u8a5e\u5f59\u672c\u8eab\u5c31\u5177\u6709\u91cd\u8907\u51fa\u73fe\u7684\u7279\u6027\uff0c\u56e0\u6b64\uff0c\uf974\u5019\u9078\u8a5e\u51fa\u73fe\u7684\u6b21\uf969\u6108\u9ad8\uff0c\u6108\u6709\u53ef\u80fd\u662f\u8a5e\u5f59\u3002 2. \u63cf\u8ff0\u9577\ufa01\u589e\u76ca(Description Length Gain, DLG) (\u516c\u5f0f ) X : \u8a9e\u6599\u5eab\u4e2d\u7684\u6240\u6709\u6587\u53e5 |X| : \u8a9e\u6599\u5eab\u4e2d\u6240\u6709\u6587\u53e5\u7684\u5b57\u5143\u7e3d\uf969 V : \u8a9e\u6599\u5eab\u4e2d\u6240\u6709\u5b57\u5143\u6240\u69cb\u6210\u7684\u96c6\u5408 L(\u3004) : \u8a9e\u6599\u5eab\u7684\u8cc7\u8a0a\uf97e(\u4e82\ufa01) X[ @ \u2192 T i ] : \u8a9e\u6599\u5eab\u6240\u6709\u6587\u53e5\u4e2d\uff0c\u5c07\u5019\u9078\u8a5e T i \u53d6\u4ee3\u6210\"@\" \u63cf\u8ff0\u9577\ufa01\u589e\u76ca\u7279\u5fb5\u662f\u7531 Kit \u7b49\u4eba\u6240\u63d0\u51fa\u4f86\u7684\u7d71\u8a08\u7279\u5fb5[7]\uff0c\u4e3b \u8981\u6982\u5ff5\u662f\u5229\u7528\u8cc7\u6599\u58d3\u7e2e\u7684\u7a0b \ufa01\u4f86\u8a55\u4f30\u5b57\u5143\u7d44\u662f\u4e00\u500b\u8a5e\u5f59\u7684\u53ef\u80fd\u6027\u3002\u516c\u5f0f 2.2 \u4e2d\u7684\uff2c(X)\u70ba\u8a9e\u6599\u5eab\u542b\u6709 T i \u7684\u8cc7\u8a0a \uf97e\uff0cL(X[@\u2192T i ])\u5247\u662f\u5c07\u8a9e\u6599\u5eab\u4e2d\u6240\u6709\u51fa\u73fe\u7684\u5019\u9078\u8a5e T i \u53d6\u4ee3\u70ba\"\uff20\"\u4e4b\u5f8c\u7684\u8cc7\u8a0a\uf97e\u3002\u56e0 \u6b64\uff0cDLG(T i )\u8868\u793a T i \u6240\u7522\u751f\u7684\u8a9e\u6599\u5eab\u8cc7\u8a0a\uf97e\u589e\u76ca\uff0c\u53ef\u4ee5\u53cd\u61c9\u51fa\u8a72\u5019\u9078\u8a5e\u5c0d\u65bc\u6574\u500b\u8a9e\u6599\u5eab \u8cc7\u8a0a\uf97e\u7684\u8ca2\u737b\ufa01\u3002\u5c0d\u8a9e\u6599\u5eab\u8cc7\u8a0a\uf97e\u8ca2\u737b\ufa01\u6108\u9ad8\u7684\u5019\u9078\u8a5e\uff0c\u6108\u53ef\u80fd\u662f\u500b\u8a5e\u5f59\u3002 3. \u4ecb\u63a5\u8b8a\u7570\ufa01(Accessor Variety, AV) AV(T i )= min{ L AV (T i ), R AV (T i )} (\u516c\u5f0f 2.3) L AV (T i ) : \u5019\u9078\u8a5e\u5de6\u908a\u76f8\u9130\u4e0d\u540c\u5b57\u5143\u7684\u500b\uf969 R AV (T i) :", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "S(T i ;k,l)\u3006\u5f9e\u5019\u9078\u8a5e T i \u4e2d\u53d6\u51fa\u4f4d\u7f6e k \u5230 l \u7684\u5b57\u5143\u7d44\u3002 \u4ee5\u5b57\u5143\u7d44\u300c\u884c\u653f\u9662\u9577\u300d\u70ba\u4f8b\uff0c\u5176\u5305\u542b\u7684\u5b50\u5b57\uf905\u9664\uf9ba\"\u884c\u653f\u9662\u9577\"\u5916\u9084\u6709\"\u884c\u653f\"\u3001\"\u884c\u653f \u9662\"\u3001\"\u653f\u9662\"\u3001\u7b49\uff0c\uf94f\u52a0\u6240\u6709\u5b50\u5b57\uf905\u7684\u51fa\u73fe\u6b21\uf969\uff0c\u5373\u53ef\u5f97\u5230\u93c8\u7d50\u5f37\ufa01\u3002 5. \u5b57\u9996\u5206\u96e2\ufa01(PreC) \u6211\u5011\u5229\u7528\u8207\u5b57\u5143\u7d44\u64c1\u6709\u76f8\u540c\u7684\u5b57\u9996\u7684\u5176\u4ed6\u5b57\u5143\u7d44\u8207\u5176\u5b50\u5b57\u5143\u7d44\u7684\u8cc7\u8a0a\uff0c\u52a0\u5f37\u6b64\u7279\u5fb5\u503c \u7684\u53ef\u9760\ufa01\uff1b\u4e5f\u5c31\u662f\u900f\u904e\u5177\u6709\u76f8\u540c\u5b57\u9996\u7684\u5b57\u5143\u7d44\uff0c\u4e00\u8d77\u8a08\u7b97\u51fa\u5b57\u9996\u7684\u5206\u96e2\u7a0b\ufa01\u3002\u5982\u679c\u5b57 \u9996\u7684\u5206\u96e2\ufa01\u6108\u5927\uff0c\u5247\u5019\u9078\u7684\u5b57\u5143\u7d44\u8f03\u53ef\u80fd\u4e0d\u662f\u8a5e\u5f59\u3002\u5176\u8a08\u7b97\u516c\u5f0f\u5982\u4e0b\u3006 (\u516c\u5f0f ) F\u3006T i \u7684\u5b57\u9996 S(F)\u3006\u4ee5 F \u5b57\u5143\u70ba\u5b57\u9996\u7684\u5b57\u5143\u7d44\u6240\u7d44\u6210\u7684\u96c6\u5408\uff0c\u4e14\u5b57\u5143\u7d44\u9577\ufa01\u9700\u5927\u65bc 2 |S(F)|\u3006S \u96c6\u5408\u7684\u5b57\u5143\u7d44\u7e3d\uf969 \u3006\u4e0d\u5305\u542b\u5b57\u9996\u7684 x \u5b50\u5b57\u5143\u7d44 x[1:L] \u6211\u5011\u91dd\u5c0d\u500b\u5225\u5b57\u9996\uff0c\u5148\u53d6\u51fa\u9577\ufa01\u70ba 3 \u5230 7 \u7684\u5b57\u5143\u7d44\uff0c\u8a08\u7b97\u5176\u79fb\u9664\u5b57\u9996\u5f8c\u5b50\u5b57\u5143\u7d44\u51fa\u73fe\u6b21 \uf969\u7684\u5e73\u5747\uff0c\u4f5c\u70ba\u8a72\u5b57\u9996\u7684\u5206\u96e2\ufa01\uff0c\uf974\u5019\u9078\u8a5e\u9577\ufa01\u5927\u65bc 2\uff0c\u8f38\u51fa\u5176\u5b57\u9996\u5206\u96e2\ufa01\uff0c\u5c0f\u65bc\u7b49\u65bc 2 \u6642\uff0c\u5247\u4ee5\u5019\u9078\u8a5e\u51fa\u73fe\u6b21\uf969\u66ff\u4ee3\u3002\u4f8b\u5982\u3006 \u300c\u5728\u53f0\u5317\u300d \uff0c\u8207\u5176\u76f8\u540c\u5b57\u9996\u7684\u5b57\u5143\u7d44\u6709 \u300c\u5728\u62cd\u651d\u300d \u3001 \u300c\u5728\u5b78\u6821\u300d\u7b49\uff0c\u5247\u5206\u5225\u7d71\u8a08\u5b50\u5b57\u5143\u7d44\u300c\u53f0\u5317\u300d \u3001 \u300c\u62cd\u651d\u300d \u3001 \u300c\u5b78\u6821\u300d\u7684\u51fa\u73fe\u6b21\uf969\uff0c\u8a08\u7b97\u5176\u5e73 \u5747\u503c\uff0c\u5373\u53ef\u53d6\u5f97\u5b57\u9996\u5206\u96e2\ufa01\u3002 \u7531\u65bc\u524d\u9762\u6240\u8ff0\u7684\u7279\u5fb5\u90fd\u662f\u7531\u8a9e\u6599\u5eab\u7d71\u8a08\u5f97\u5230\uff0c\u9019\u4e9b\u7279\u5fb5\u7684\uf969\u503c\u6703\u53d7\u8a9e\u6599\u5eab\u5927\u5c0f\u7684\u5f71 \u97ff\uff0c\u800c\u843d\u5728\u4e0d\u540c\u7684\u7bc4\u570d\u3005\u5982\u679c\u6e2c\u8a74\u548c\u8a13\u7df4\u7684\u8a9e\u6599\u5eab\u5927\u5c0f\u6709\u660e\u986f\u7684\u5dee\u7570\u6642\uff0c\u5247\u8a13\u7df4\u548c\u6e2c\u8a74 \u7684\u7d71\u8a08\u7279\u5fb5\u503c\u5c07\u843d\u5728\u4e0d\u540c\u7684\u52d5\u614b\u7bc4\u570d\uff0c\u9019\u4f7f\u5f97\u8a13\u7df4\u51fa\u4f86\u7684\u5206\u985e\u5668\u7121\u6cd5\u5c0d\u6e2c\u8a74\u7684\u8cc7\u6599\u505a\u53ef \u9760\u5730\u5206\u985e\u3002\u70ba\uf9ba\u89e3\u6c7a\u9019\u500b\u554f\u984c\uff0c\u6211\u5011\u628a\u4e0a\u8ff0\u7684\u5404\u500b\u7d71\u8a08\u7279\u5fb5\u4ee5\u7dda\u6027\u65b9\u5f0f\u6b63\u898f\u5316\u52300\u81f31 \u4e4b\u9593\uff0c\u516c\u5f0f\u5982\u4e0b\u3006 (\u516c\u5f0f ) v : \u8f38\u5165\u7684\u7279\u5fb5\uf969\u503c : \u7279\u5fb5\u503c\u7684\u7a2e\u985e Min( ) : \u7279\u5fb5\u503c\u4e2d\u6700\u5c0f\u7684\uf969\u503c Max( ) : \u7279\u5fb5\u503c\u4e2d\u6700\u5927\u7684\uf969\u503c F(v) : \u7279\u5fb5\u503c v", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "X MAX \u5716 3.2 HEQ \u6b63\u898f\u5316\u65b9\u6cd5\u793a\u610f\u5716 y X d X S \u2500\u2500 P(X) ---- P EQ (X) P EQ (X d ) P(X S ) X MIN 1 X \u5c0d\u6240\u6709\u5019\u9078\u8a5e\u8a08\u7b97 LogC\u3001AV", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "", | |
| "sec_num": null | |
| } | |
| ], | |
| "back_matter": [], | |
| "bib_entries": { | |
| "BIBREF0": { | |
| "ref_id": "b0", | |
| "title": "Identifying Chinese Name in Unrestricted Texts", | |
| "authors": [ | |
| { | |
| "first": "", | |
| "middle": [], | |
| "last": "Ms Sun", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "", | |
| "middle": [], | |
| "last": "Hung", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Jfang", | |
| "middle": [], | |
| "last": "Gao", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1994, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "MS Sun, CN Hung, HY Gao, and JFang, \"Identifying Chinese Name in Unrestricted Texts\", Chinese & Oriental Languages Information Processing Society, 1994.", | |
| "links": null | |
| }, | |
| "BIBREF1": { | |
| "ref_id": "b1", | |
| "title": "Statistical Substring Reduction in Linear Time", | |
| "authors": [ | |
| { | |
| "first": "Xueqiang", | |
| "middle": [], | |
| "last": "Chunyu", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Le", | |
| "middle": [], | |
| "last": "Lu", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Junfeng", | |
| "middle": [], | |
| "last": "Zhang", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "", | |
| "middle": [], | |
| "last": "Hu", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2004, | |
| "venue": "Proceeding of the 1nd International Joint Conference on Natural Language Processing(IJCNLP)", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Chunyu,Xueqiang Lu, Le Zhang, and Junfeng Hu, \"Statistical Substring Reduction in Linear Time\", In Proceeding of the 1nd International Joint Conference on Natural Language Processing(IJCNLP), 2004.", | |
| "links": null | |
| }, | |
| "BIBREF2": { | |
| "ref_id": "b2", | |
| "title": "An Empirical Comparison of Goodness Measures for Unsupervised Chinese Word Segmentation with a Unified Framework", | |
| "authors": [ | |
| { | |
| "first": "Zhao", | |
| "middle": [], | |
| "last": "Hai", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Kit", | |
| "middle": [], | |
| "last": "Chunyu", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2008, | |
| "venue": "Proceedings of The 3nd International Joint Conference on Natural Language Processing(IJCNLP)", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Zhao Hai and Kit Chunyu, \"An Empirical Comparison of Goodness Measures for Unsupervised Chinese Word Segmentation with a Unified Framework\", In Proceedings of The 3nd International Joint Conference on Natural Language Processing(IJCNLP), 2008.", | |
| "links": null | |
| }, | |
| "BIBREF3": { | |
| "ref_id": "b3", | |
| "title": "Unknown Word Extraction for Chinese Documents", | |
| "authors": [ | |
| { | |
| "first": "Keh-Jiann", | |
| "middle": [], | |
| "last": "Chen", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Wei-Yun", | |
| "middle": [], | |
| "last": "Ma", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2002, | |
| "venue": "Proceedings of The 19nd International Conference on Computational Linguistics (COLING)", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "169--175", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Keh-Jiann Chen and Wei-Yun Ma, \"Unknown Word Extraction for Chinese Documents\", In Proceedings of The 19nd International Conference on Computational Linguistics (COLING), Pages 169-175, 2002.", | |
| "links": null | |
| }, | |
| "BIBREF4": { | |
| "ref_id": "b4", | |
| "title": "Proceedings of Research on Computational Linguistics Conference XIII(ROCLING)", | |
| "authors": [ | |
| { | |
| "first": "\uf96e\u5927\u69ae", | |
| "middle": [], | |
| "last": "\uf97a\u5a77", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "", | |
| "middle": [], | |
| "last": "\u61c9\u7528\u69cb\u8a5e\u6cd5\u5247\u8207\u985e\u795e\u7d93\u7db2\u8def\u65bc\u4e2d\u6587\u65b0\u8a5e\u8403\u53d6", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2000, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "21--40", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "\uf97a\u5a77, \uf96e\u5927\u69ae, \u61c9\u7528\u69cb\u8a5e\u6cd5\u5247\u8207\u985e\u795e\u7d93\u7db2\u8def\u65bc\u4e2d\u6587\u65b0\u8a5e\u8403\u53d6, In Proceedings of Research on Computational Linguistics Conference XIII(ROCLING), Pages 21-40, 2000.", | |
| "links": null | |
| }, | |
| "BIBREF5": { | |
| "ref_id": "b5", | |
| "title": "Chinese unknown word identification using character-based tagging and chunking", | |
| "authors": [ | |
| { | |
| "first": "Masayuki", | |
| "middle": [], | |
| "last": "Goh Chooi Ling", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Yuji", | |
| "middle": [], | |
| "last": "Asahara", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "", | |
| "middle": [], | |
| "last": "Matsumoto", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2003, | |
| "venue": "Proceedings of The 41nd Annual Meeting on Association for Computational Linguistics", | |
| "volume": "2", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Goh Chooi Ling, Masayuki Asahara, and Yuji Matsumoto, \"Chinese unknown word identification using character-based tagging and chunking\", In Proceedings of The 41nd Annual Meeting on Association for Computational Linguistics -Volume 2, Pages 197-200, 2003.", | |
| "links": null | |
| }, | |
| "BIBREF6": { | |
| "ref_id": "b6", | |
| "title": "Unsupervised Learning of Word Boundary with Description Length Gain", | |
| "authors": [ | |
| { | |
| "first": "Chunyu", | |
| "middle": [], | |
| "last": "Kit", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Yorick", | |
| "middle": [], | |
| "last": "Wilks", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1999, | |
| "venue": "Proceedings of CoNLL99 ACL Workshop", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Chunyu Kit and Yorick Wilks, \"Unsupervised Learning of Word Boundary with Description Length Gain\", In Proceedings of CoNLL99 ACL Workshop, 1999.", | |
| "links": null | |
| }, | |
| "BIBREF7": { | |
| "ref_id": "b7", | |
| "title": "Accessor Variety Criteria for Chinese Word Extraction", | |
| "authors": [ | |
| { | |
| "first": "Haodi", | |
| "middle": [], | |
| "last": "Feng", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Kang", | |
| "middle": [], | |
| "last": "Chen", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Xiaotie", | |
| "middle": [], | |
| "last": "Deng", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Weimin", | |
| "middle": [], | |
| "last": "Zheng", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2004, | |
| "venue": "Computational Linguistics", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Haodi Feng, Kang Chen, Xiaotie Deng, and Weimin Zheng, \"Accessor Variety Criteria for Chinese Word Extraction\", Computational Linguistics, 2004.", | |
| "links": null | |
| }, | |
| "BIBREF8": { | |
| "ref_id": "b8", | |
| "title": "Chinese Terminology Extraction Using Window-Based Contextual Information", | |
| "authors": [ | |
| { | |
| "first": "Luning", | |
| "middle": [], | |
| "last": "Ji", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Mantai", | |
| "middle": [], | |
| "last": "Sum", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Qin", | |
| "middle": [], | |
| "last": "Lu", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Wenjie", | |
| "middle": [], | |
| "last": "Li", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Yirong", | |
| "middle": [], | |
| "last": "Chen", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2009, | |
| "venue": "Proceedings of the 8th International Conference on Computational Linguistics and Intelligent Text Processing", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "62--74", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Luning Ji, Mantai Sum, Qin Lu, Wenjie Li, and Yirong Chen, \"Chinese Terminology Extraction Using Window-Based Contextual Information\", In Proceedings of the 8th International Conference on Computational Linguistics and Intelligent Text Processing, Pages: 62 -74, 2009", | |
| "links": null | |
| }, | |
| "BIBREF9": { | |
| "ref_id": "b9", | |
| "title": "Image enhancement by histogram transformation", | |
| "authors": [ | |
| { | |
| "first": "Robert", | |
| "middle": [], | |
| "last": "Hummel", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1977, | |
| "venue": "Comp. Graph. Image Process", | |
| "volume": "6", | |
| "issue": "", | |
| "pages": "184--195", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Robert Hummel,\" Image enhancement by histogram transformation\", Comp. Graph. Image Process. , vol. 6, Pages 184-195, 1977.", | |
| "links": null | |
| } | |
| }, | |
| "ref_entries": { | |
| "TABREF2": { | |
| "type_str": "table", | |
| "content": "<table><tr><td/><td>DLG\u7279\u5fb5\u503c\u5206\u4f48 P EQ (X)\u3006\u5747\u5316\u5206\u4f48\u4e4b\uf94f\u7a4d\u5206\u4f48\u51fd\uf969(CDF)</td><td>DLG\u7279\u5fb5\u503c\u5206\u4f48</td></tr><tr><td>60%</td><td>X max \u3006\u7279\u5fb5\u7684\u6700\u5927\u503c</td><td>60%</td></tr><tr><td colspan=\"3\">(\u4e8c)\u3001\u5206\u4f48\u6b63\u898f\u5316\u7684\u6cd5\u65b9\u4ecb\u7d39 \u8a13\u7df4\u8207\u6e2c\u8a74\u7279\u5fb5\u503c\u5206\u4f48\u6709\u660e\u986f\u5dee\u7570\u6642\uff0c\u8a13\u7df4\u51fa\u4f86\u7684\u5206\u985e\u5668\u5c31\u7121\u6cd5\u53ef\u9760\u5730\u5206\u985e\u6e2c\u8a74\u9818\u57df\u7684 \u8cc7\u6599\u3002\u7531\u65bc\u6211\u5011\u5e0c\u671b\u8a5e\u5f59\u8403\u53d6\u6280\u8853\u53ef\u4ee5\u7528\u4f86\u63a2\u7d22\u672a\u77e5\u7684\u65b0\u9818\u57df\uff0c\u5728\u65b0\u9818\u57df\u4e2d\u7279\u5fb5\u7684\u7d71\u8a08 \u5206\u4f48\u5f88\u53ef\u80fd\u4e0d\u540c\u65bc\u8a13\u7df4\u9818\u57df\uff0c\u56e0\u6b64\u5fc5\u9808\u514b\u670d\u6b64\u554f\u984c\u3002\u672c\u8ad6\u6587\u5206\u5225\u70ba\u4f7f\u7528\u6a19\u6e96\u5dee\u500d\uf969\u6cd5\u8207 \u76f4\u65b9\u5716\u5747\u5316\u6cd5\u9032\u4e00\u6b65\u6b63\u898f\u5316 DLG \uf969\u503c\u3002\u9996\u5148\u4ecb\u7d39\u6a19\u6e96\u5dee\u500d\uf969\u6b63\u898f\u5316\u65b9\u6cd5(Mean -Standard Deviation Weight, \u7c21\u7a31 MSW) \uff0c\u516c\u5f0f\u5982\u4e0b\u3006 \uff44 (\u516c\u5f0f d : destination \u76ee\u6a19\u9818\u57df S : source \u4f86\u6e90\u9818\u57df M d \u3006\u76ee\u6a19\u9818\u57df(\u8a13\u7df4\u8cc7\u6599)\u7279\u5fb5\u5206\u4f48\u4e2d\u7684\u5e73\u5747\uf969 M S \u3006\u4f86\u6e90\u9818\u57df(\u6e2c\u8a74\u8cc7\u6599)\u7279\u5fb5\u5206\u4f48\u4e2d\u7684\u5e73\u5747\uf969 \u03c3 d \u3006\u76ee\u6a19\u9818\u57df\u7279\u5fb5\u5206\u4f48\u4e2d\u7684\u6a19\u6e96\u5dee \u03c3 S \u3006\u4f86\u6e90\u9818\u57df\u7279\u5fb5\u5206\u4f48\u4e2d\u7684\u6a19\u6e96\u5dee 0% 20% 40% 0.04 0.049 0.058 0.067 0.076 0.085 0.094 CKIP_Train CKIP_Test \u8cc7 \u6599 \u91cf 0% 20% 40% 0.02 0.032 0.044 0.056 0.068 0.08 0.092 CKIP_Train HKCU_Test \u8cc7 X min \u3006\u7279\u5fb5\u7684\u6700\u5c0f\u503c \u6599 \u91cf \u5716 3.1 \u4e0d\u540c\u8a9e\u6599\u5eab\u4e4b DLG \u7279\u5fb5\u503c\u5206\u4f48\u6bd4\u8f03\u5716 (b) \u8de8\u9818\u57df\u5206\u4f48\u5716 (a) \u76f8\u540c\u9818\u57df\u5206\u4f48\u5716 \u5716 3.2 \u63db\u81f3\u8a13\u7df4\u8cc7\u6599\u7684\u7279\u5fb5\u7a7a\u9593\u3002\u5728\u9032\u884c\u76f4\u65b9\u5716\u5747\u5316\u6b63\u898f\u5316\u6642\uff0c\u5247\u662f\u8a13\u7df4\u8cc7\u6599\u548c\u6e2c\u8a74\u8cc7\u6599\u5747\u9808</td></tr><tr><td colspan=\"3\">X S \u3006\u4f86\u6e90\u9818\u57df\u7684\u7279\u5fb5\u503c \u6839\u64da\u516c\u5f0f 3.2 \u8f49\u63db\uff0c\u4f46\u8f49\u63db\u6240\u4f7f\u7528\u7684 CDF \u662f\u5206\u5225\u7531\u8a13\u7df4\u8cc7\u6599\u8207\u6e2c\u8a74\u8cc7\u6599\u7d71\u8a08\u5f97\u5230\u3002\u6240</td></tr><tr><td colspan=\"3\">X d \u3006\u8f49\u63db\u5f8c\u7684\u76ee\u6a19\u9818\u57df\u7279\u5fb5\u503c \u4ee5\u6a19\u6e96\u5dee\u500d\uf969\u6cd5\u662f\u5c07\u6e2c\u8a74\u8cc7\u6599\u7684\u7279\u5fb5\u7a7a\u9593\u8f49\u63db\u81f3\u8a13\u7df4\u8cc7\u6599\u7684\u7279\u5fb5\u7a7a\u9593\u3005\u76f4\u65b9\u5716\u5747\u5316\u6cd5\u5247</td></tr><tr><td colspan=\"3\">\u6a19\u6e96\u5dee\u500d\uf969\u6b63\u898f\u5316\u662f\u4e00\u7a2e\u7dda\u6027\u8abf\u6574\u7684\u65b9\u6cd5\u3002\u8f49\u63db\u65b9\u5f0f\u662f\u4ee5\u4f86\u6e90\u9818\u57df(\u6e2c\u8a74\u8cc7\u6599)\u6a19\u6e96\u5dee \u662f\u5c07\u6e2c\u8a74\u8cc7\u6599\u8207\u8a13\u7df4\u8cc7\u6599\u7684\u7279\u5fb5\u7a7a\u9593\uff0c\u540c\u6642\u8f49\u63db\u81f3\u7dda\u6027\u7684 CDF \u5206\u4f48\u7a7a\u9593\u3002</td></tr><tr><td colspan=\"3\">\u70ba\u8861\uf97e\u57fa\u6e96\u55ae\u4f4d\uff0c\u8a08\u7b97\u7279\u5fb5\uf969\u503c\u8207\u5176\u5206\u4f48\u5e73\u5747\u503c\u4e4b\u9593\u7684\u6b63\u898f\u5316\u8ddd\u96e2\u70ba\u591a\u5c11\u500d\u6a19\u6e96\u5dee\uff0c\u518d (\u4e09)\u3001\u985e\u795e\u7d93\u5206\u985e\u5668</td></tr><tr><td colspan=\"3\">\u76f8\u540c\u9818\u57df\u4e14\u8cc7\u6599\uf97e\u76f8\u8fd1\u7684\u8a9e\u6599\u5eab\uff0c\u7d71\u8a08\u51fa\u4f86\u7684 DLG \u7279\u5fb5\u503c\u5206\u4f48\uff0c\u5f7c\u6b64\u53ea\u6709\u4e9b\u5fae\u5dee\u7570\uff0c\u4f46 \u63db\u7b97\u6210\u76ee\u6a19\u9818\u57df(\u8a13\u7df4\u8cc7\u6599)\u7684\u503c\u3002 \u8a5e\u5f59\u7d44\u6210\u7684\u7d50\u69cb\u548c\u4f7f\u7528\u65b9\u5f0f\u975e\u5e38\u8907\u96dc\uff0c\u55ae\u9760\u4e00\u7a2e\u7279\u5fb5\u503c\u901a\u5e38\u4e26\u4e0d\u80fd\u505a\u53ef\u9760\u7684\u5224\u65b7\uff0c\u5f80\u5f80 \u63a5\u8457\u4ecb\u7d39\u76f4\u65b9\u5716\u5747\u5316\u6cd5(Histogram Equalization, \u7c21\u7a31 HEQ)[9]\uff0c\u5176\u8f49\u63db\u51fd\u5f0f\u5982\u4e0b\u3006 \u5fc5\u9808\u7d50\u5408\u591a\u7a2e\u7279\u5fb5\uff0c\u4ee5\u767c\u63ee\u4e92\u88dc\u7684\u529f\u6548\u3002\u5229\u7528\u5206\u985e\u5668\u53ef\u4ee5\u5f48\u6027\u5730\u7d50\u5408\u4e0d\u540c\u7684\u7279\u5fb5\uff0c\u4ee5\u9054 \u662f\u5728\u5716 3.1(b)\u4e2d\uff0c\u56e0\u70ba\u4e0d\u540c\u9818\u57df\u7684\u8a9e\u6599\u5eab\u3001\u8cc7\u6599\uf97e\u660e\u986f\u7684\u5dee\u7570\u7b49\u539f\u56e0\uff0c\u4f7f\u5f97 DLG \u7279\u5fb5 X d \uff1d P(X S )\u3003( X max -X min ) + X min (\u516c\u5f0f 3.2) \u5230\u66f4\u597d\u7684\u5206\u985e\u6548\u679c\u3002\u56e0\u6b64\uff0c\u672c\u8ad6\u6587\u4e2d\u4f7f\u7528\uf9ba\u591a\u5c64\u6b21\u985e\u795e\u7d93\u7db2\u8def\u5206\u985e\u5668\uff0c\u9032\u884c\u8a5e\u5f59\u7684\u9a57\u8b49\u3002 \u503c\u5206\u4f48\u6709\u660e\u986f\u7684\u5dee\u7570\u3002\u56e0\u6b64\uff0c\u4f7f\u7528 CKIP_Train \u8a13\u7df4\u51fa\u4f86\u7684\u5206\u985e\u5668\uff0c\u5c1a\u53ef\u88ab\u61c9\u7528\u65bc\u5206\u985e X S \u3006\u7279\u5fb5\u503c \u9019\u662f\u4e00\u7a2e\u56de\u6b78\u7684\u65b9\u6cd5\uff0c\u53ef\u4ee5\u5be6\u73fe\u975e\u7dda\u6027\u7684\u5206\u985e\u3002\u5b83\u7684\u5b78\u7fd2\u904e\u7a0b\u662f\u4ee5\u932f\u8aa4\u7684\u5012\u50b3\u905e\u65b9\u5f0f\uff0c\u91cd CKIP_Test \u7684\u8a9e\u6599\u3005\u4f46\u662f\u61c9\u7528\u65bc\u8de8\u9818\u57df\u7684\u8cc7\u6599\u6642\uff0c\u5176\u7d71\u8a08\u5206\u4f48\u6709\u660e\u986f\u7684\u5dee\u7570\uff0c\u5c0e\u81f4\u6b64\u5206 \u985e\u5668\u7121\u6cd5\u53ef\u9760\u5730\u5206\u985e\u3002\u6240\u4ee5\u6211\u5011\u91dd\u5c0d DLG \u7279\u5fb5\u503c\u9032\u4e00\u6b65\u505a\u6b63\u898f\u5316\uff0c\u4f7f\u5176\u8a13\u7df4\u8cc7\u6599\u8207\u6e2c X d \u3006\u5747\u5316\u5f8c\u7684\uf969\u503c \u8907\u8fed\u4ee3\u7db2\u8def\u6b0a\u91cd\u503c\uff0c\u4f7f\u5f97\u7e3d\u5e73\u65b9\u8aa4\u5dee\u6700\u5c0f\u5316\u3002</td></tr><tr><td colspan=\"3\">\u8a74\u8cc7\u6599\u4e2d\u7684\u7d71\u8a08\u5206\u4f48\uff0c\u53ef\u4ee5\u4e92\u76f8\u5339\u914d\u3002 P(X S )\uff1a\u7279\u5fb5\u503c\u4e4b\uf94f\u7a4d\u5206\u4f48\u51fd\uf969(CDF) \u6211\u5011\u4f7f\u7528\u985e\u795e\u7d93\u7db2\u8def\u5206\u985e\u5668\u9032\u884c\u8a5e\u5f59\u9a57\u8b49\u7684\u67b6\u69cb\u5982\u5716 3.3 \u6240\u793a\u3002\u4ee5\u516c\u5f0f(2.1)~\u516c\u5f0f(2.5)</td></tr></table>", | |
| "html": null, | |
| "text": "CKIP_Test \u4e2d\u7684 DLG \u5206\u4f48\u5dee\u7570\uff0c\u5716(b)\u5247\u986f\u793a\u8de8\u9818\u57df\u8a9e\u6599\u5eab CKIP_Train \u548c HKCU_Test \u4e2d\u7684 DLG \u5206\u4f48\u5dee\u7570\u3002\u5206\u5225\u4ee5\u8a9e\u6599\u5eab\u7684\u540d\u7a31\u547d\u540d\u5206\u4f48\u66f2\u7dda\uff0c\u4f8b\u5982\u3006\u5716(a)\u4e2d\u7684 CKIP_Train \u66f2\u7dda\uff0c\u8868\u793a\u5f9e CKIP_Train \u8a9e\u6599\u5eab\u4e2d\u7d71\u8a08\u5f97\u5230\u7684 DLG \u7279\u5fb5\u503c\u5206\u4f48\u3002\u7531\u5716 3.1(a)\u53ef\u4ee5\u770b\u51fa\uff0c\u5f9e \u662f HEQ \u6b63\u898f\u5316\u65b9\u6cd5\u7684\u793a\u610f\u5716\uff0c\u7e31\u8ef8\u70ba\u7279\u5fb5 X \u7684\uf94f\u7a4d\u5206\u4f48\u51fd\uf969(Cumulative Distribution Function, CDF) \uff0c\u6a6b\u8ef8\u70ba\u7279\u5fb5 X \u7684\u503c\u3002\u6b64\u8f49\u63db\u65b9\u5f0f\u662f\u5229\u7528\uf94f\u7a4d\u5206\u4f48\u51fd\uf969\uff0c\u5c07 X S \u8f49\u63db\u5230\u5747\u5316\u5206\u4f48 P EQ (X)\u4e0a\u5177\u6709\u76f8\u540c CDF \u503c\u7684\u7279\u5fb5 X d \uff0c\u4e5f\u5c31\u662f\u5c07 X \u7279\u5fb5\u7684\u4f30\u6e2c CDF \u5206\u4f48\u6620\u5c04\u5230\u7dda\u6027 CDF \u5206\u4f48\u7684\u7a7a\u9593\u4e2d\u3002\u800c\u7dda\u6027 CDF \u6240\u5c0d\u61c9\u7684\u6a5f\uf961\u5bc6\ufa01\u51fd\uf969\u503c (Probability Density Function, PDF)\u662f\u4e00\u5747\u52fb\u5206\u4f48(uniform distribution) \uff0c\u6545\u7a31\u70ba\u300c\u5747\u5316\u300d \u3002\u76f4\u65b9\u5716\u5747 \u5316\u662f\u4e00\u7a2e\u55ae\u8abf(monotonic)\u7684\u8f49\u63db\u65b9\u5f0f\uff0c\u6839\u64da\u7279\u5fb5\u503c\u7684\u8cc7\u6599\uf97e\u505a\u975e\u7dda\u6027\u8abf\u6574\uff0c\u8abf\u6574\u5f8c \u7684\uf969\u503c\u80fd\u5e73\u5747\u5206\u4f48\u65bc\u76f8\u540c\u52d5\u614b\u7bc4\u570d\u4e2d(X min \u5230 X max \u4e4b\u9593) \u3002\u4f7f\u7528 HEQ \u6b63\u898f\u5316\u65b9\u6cd5\u6642\uff0c\u5fc5 \u9808\u5c07\u8a13\u7df4\u7279\u5fb5\u548c\u6e2c\u8a74\u7279\u5fb5\u90fd\u5fc5\u9808\u4f7f\u7528\u516c\u5f0f 3.2 \u9032\u884c\u5747\u5316\uff0c\u4f7f\u7d71\u8a08\u5206\u4f48\u540c\u6642\u8f49\u63db\u81f3\u7dda\u6027 CDF \u5206\u4f48\u7a7a\u9593\uff0c\u8b93\u5f7c\u6b64\u53ef\u4ee5\u4e92\u76f8\u5339\u914d\uff0c\u89e3\u6c7a\u7d71\u8a08\u5206\u4f48\u5dee\u7570\u9020\u6210\u5206\u985e\u5668\u7121\u6cd5\u53ef\u9760\u5730\u5206\u985e \u554f\u984c\u3002 \u5728\u9032\u884c\u6a19\u6e96\u5dee\u500d\uf969\u6b63\u898f\u5316\u524d\uff0c\u9808\u5148\u5c0d\u8a13\u7df4\u8cc7\u6599\u7d71\u8a08 M d \u3001\u03c3 d \u53ca\u5c0d\u6e2c\u8a74\u8cc7\u6599\u8a08\u7b97 M S \u3001\u03c3 S \u3002 \u8a13\u7df4\u8cc7\u6599\u8a08\u7b97\u51fa\u7684 DLG \u7279\u5fb5\u4e0d\u9808\u505a\u8f49\u63db\uff0c\u4f46\u6e2c\u8a74\u8cc7\u6599\u4e4b DLG \u7279\u5fb5\u9808\u6839\u64da\u516c\u5f0f 3.1\uff0c\u8f49", | |
| "num": null | |
| }, | |
| "TABREF3": { | |
| "type_str": "table", | |
| "content": "<table><tr><td>\u53ef\u70ba\u55ae\u4e00\u7279\u5fb5\u6216\u591a\u7dad\u7279\u5fb5\u3002\u5206\u985e\u5668\u7684\u8f38\u51fa y \u70ba 0 \u5230 1 \u4e4b\u9593\u7684\uf969\u503c\uff0c\u7d93\u9580\u6abb\u6e2c\u8a74\uff0c\u9032\u884c\u8a5e \u5f59\u9a57\u8b49\uff0c\u6c7a\u7b56\u5019\u9078\u8a5e\u662f\u5426\u70ba\u4e00\u8a5e\u5f59\u3002 \u662f \u5426 \u8a5e\u5f59 \u975e\u8a5e\u5f59 x y HEQ/MSW \u7279\u5fb5\u9078\u53d6 F(\u2027) F(\u2027) F(\u2027) F(\u2027) F(\u2027) DLG AV Link PreC LogC \u5012\u50b3\u905e \u5b78\u7fd2 \u6b0a\u91cd \u6839\u64da\u4e0a\u9762\u7d50\u679c\u7684\u7d50\u679c\uff0c\u6211\u5011\u767c\u73fe HEQ \u53ef\u4ee5\u8b93\u8a13\u7df4\u548c\u6e2c\u8a74\u7684\u7d71\u8a08\u5206\u4f48\u4e92\u76f8\u5339\u914d\uff0c\u61c9\u7528\u5728 \u8de8\u9818\u57df\u7684\u8cc7\u6599\uff0c\u4f7f\u5f97\u8a13\u7df4\u6a21\u578b\u66f4\u52a0\u4e00\u822c\u5316\u3002\u4e0d\u50c5\u5982\u6b64\uff0c\u4e5f\u8868\u793a DLG \u7279\u5fb5\u503c\u9069\u5408\u5229\u7528 HEQ \u5716 4.2 \u8de8\u9818\u57df\u4e4b\u4e0d\u540c\u6b63\u898f\u5316\u65b9\u5f0f\u7684\u6548\u80fd\u6bd4\u8f03\u5716 \u908a\u5713\u578b\u8868\u793a\u5019\u9078\u8a5e\u96c6\u5b58\u5728\u65bc CKIP_Dic \u7684\u8a5e\u5f59\u96c6\u5408 (\u7c21\u7a31 CKIP_Words) \uff0c\u7e3d\u5171\u5305\u542b 11,290 \u90a3\u746a\u590f\u9109 \u528d\u6e56\u5c71 \u7fa9\u8ce3\u54c1 \u99ac\u7e3d\u7d71 \u5f35\u745e\u8ce2 \u7d93\u767c\u5c40 \u672a\u77e5\u8a5e\u96c6\u5408\uff0c\u5171\u6709 2,404 \u500b\u672a\u77e5\u8a5e\uff0c\u672c\u8ad6\u65b9\u6cd5\u62bd\u51fa\u4f86\u7684\u672a\u77e5\u8a5e\uff0c\u6709 911 \u500b\u76f8\u540c\u672a\u77e5\u8a5e\u5f59\u3002 \u8a5e\u5178(CKIP_Dic \u8207 ICTCAS_Dic)\u9032\u884c\u8a5e\u5f59\u7684\u6a19\u8a18\uff0c\u6a19\u8a18\u7684\u7d50\u679c\u5982\u5716 5.1(a)\u6240\u793a\u3002\u5716\u4e2d\u53f3 \u62bd\u53d6\u7684 1,486 \u500b\u8a5e\u4e2d\uff0c\u6709 567 \u500b\u76f8\u540c\u7684\u8a5e\u5f59\u3002\u5716 5.2(b)\u4e2d\u5247\u770b\u5230 CKIP \u65b7\u8a5e\u7cfb\u7d71\u62bd\u51fa\u7684 \u8a9e\u97f3\u4fe1\u7bb1 \u91d1\u878d\u6d77\u562f \u7e3d\u57f7\u884c\u9577 \u6cf0\u6b66\u6751 \u9802\u5471\u5471 \u526f\u99d5\u99db \u63a5\u4e0b\u4f86\u6211\u5011\u91dd\u5c0d\u5f9e UKW_Test \u8a9e\u6599\u5eab\u4e2d\u521d\u6b65\u7be9\u9078\u7684\u5019\u9078\u8a5e\u96c6\uff0c\u5404\u5225\u5229\u7528\u4e0a\u8ff0\u7522\u751f\u7684\u5169\u500b \u7d10\u7d04 \u5e02\u9577 \u7d10\u7d04\u5e02\u9577 \u7d10\u7d04\u5e02\u9577\u9b6f\u8fea \u6731\u5229\u5b89\u5c3c\u7576\u4e0a\u7d10\u7d04\u5e02\u9577\u5f8c 5.4(a)\u4e2d\u53ef\u770b\u5230\uff0cICTCAS \u65b7\u8a5e\u7cfb\u7d71\u62bd\u51fa\u7684\u672a\u77e5\u8a5e\u96c6\u5408\uff0c\u5171\u6709 1,477 \u500b\u672a\u77e5\u8a5e\uff0c\u800c\u672c\u8ad6\u6587 \u7d19\u6559\u5802 \u9f8d\u773c\u4e7e \u5149\uf9f4\u6751 \u6885\u5c71\u9109 \u79cb\u7bc0\u79ae\u54c1 \u6bc0\u65bc\u4e00\u65e6 \u63a5\u8457\u6211\u5011\u6bd4\u8f03\u672c\u8ad6\u6587\u65b9\u6cd5\u8207\u5169\u65b7\u8a5e\u7cfb\u7d71\u5c0d\u65bc\u8403\u53d6\u672a\u77e5\u8a5e\u7684\u5dee\u7570\u3002\u5716\u5982 5.2 \u6240\u793a\u3002\u5716 \u7db2\u8def \u985e\u795e\u7d93\u7db2\u8def \u5206\u985e\u5668 \u9580\u6abb\u6e2c\u8a74 y > \u03b7 ? \u7279\u5fb5\u7d44\u5408 F-Measure No_LogC 60.03% No_DLG 57.21% No_AV 51.74% No_Link 48.06% No_PreC 53.19% ALL \u53ec\u56de\u7387 \u8eab \u9677 \u8eab\u9677 \u881f\u7b46\u5c0f\u65b0 \u7126\u7cd6\u54e5\u54e5 \u5357\u8ff4\u516c\u8def \u6d88\u8cbb\u5238 \u8607\u7e23\u9577 \u6b63\u5927\u5149\u660e \u5247\u53ef\u80fd\u8eab\u9677\u5176\u4e2d\u7121\u6cd5\u81ea\u62d4 \u8eab\u9677\u9003\u5175\u919c\u805e\u7684\u97d3\u661f\u5b8b\u627f\u61b2 59.69% \u65b7\u8a5e\u7cfb\u7d71\u7522\u751f\u7684\u8a5e\u5178\uff0c\u5373\u70ba\u9019\u5169\u5957\u7cfb\u7d71\u5404\u5225\u8403\u53d6\u51fa\u7684\u8a5e\u5f59\u3002\u9019\u5169\u500b\u8a5e\u5178\u5206\u5225\u7a31\u70ba \u8868 5.3 \u4e09\u7a2e\u65b9\u5f0f\u7684\u8403\u53d6\u672a\u77e5\u8a5e\u7bc4\u4f8b \u8868 4.2 \u56db\u7a2e\u4ee5\u4e0a\u7279\u5fb5\u4e4b\u8a5e\u5f59\u9a57\u8b49\u6548\u80fd\u8868 0 0.10.20.30.40.50.60.70.80.9 1 \u5225 \u7121 \u9078\u64c7 \u5225\u7121\u9078\u64c7 \u90a3\u81ea\u7136\u5225\u7121\u9078\u64c7 \u6df7 \u65e5\u5b50 \u6df7\u65e5\u5b50 \u61f6\u61f6\u6563\u6563\u7684\u6df7\u65e5\u5b50 \u4ee5\u505a\u8089\uf905\u6df7\u65e5\u5b50 ICTCAS \u65b7\u8a5e\u7cfb\u7d71 10,642 \u4f73\u66ae\u82f1\u96c4 \u7da0\u8c46\u692a \u5922\u5de5\u5834 \u7c21\u5fd7\u5fe0 \u99ac\u653f\u5e9c \u76e3\u5bdf\u9662\u9577 1,477 \u9664\u6b64\u5225\u7121\u9078\u64c7 0.1 0.2 0.3 0.4 0.5 0.6 0.7 \u7cbe MSW \u5976\u7c89 \u9322 \u5976\u7c89\u9322 \u5976\u7c89\u9322\u4e5f\u6709\u9ede\u9700\u8981 CKIP \u65b7\u8a5e\u7cfb\u7d71 11,290 2,402 \u5c0f\u5de8\u86cb \u6279\u8e22\u8e22 \uf9f4\u653f\u52a9 \u4e8c\u624b\u8863 \u5e73\u5b89\u7c73 \u79c0\u59d1\u5dd2\u6eaa \u70ba\uf9ba\u8cfa\u5976\u7c89\u9322\u548c\u6559\u80b2\u57fa\u91d1 \u7387 HEQ Ckip HKCU Ckip HKCU \u672c\u8ad6\u6587\u8403\u53d6\u65b9\u6cd5 10,000 1,486 \u6d77\u89d2 7 \u865f \u529f\u592b\u704c\u7c43 \u9673\u6dfb\u52dd \u65b0\u767c\u5927\u6a4b \u6551\u96e3\u968a \u51f1\u9054\u683c\u862d \u78ba No_Equ \u8a5e\u5f59 \u53e5\u5b50 \u65b9\u6cd5 \u8a5e\u5f59\u8403\u53d6\uf969 \u672a\u77e5\u8a5e\uf969 \u672a\u77e5\u8a5e \u672a\u77e5\u8a5e \u672a\u77e5\u8a5e 0.8 \u8868 5.1 CKIP \u8207 HKCU \u4e0d\u540c\u8a5e\u5f59\u5b9a\u7fa9\u7684\u7bc4\u4f8b\u8868 0.9 1 CKIP_Dic \u8207 ICTCAS_Dic\u3002 \u8868 5.2 \u5404\u65b9\u6cd5\u4e4b\u8403\u53d6\u672a\u77e5\u8a5e\uf969\u8868 (b) ICTCAS \u65b7\u8a5e\u7cfb\u7d71 (c) CKIP \u65b7\u8a5e\u7cfb\u7d71 (a) \u672c\u8ad6\u6587\u65b9\u6cd5</td></tr><tr><td>\u5c0d\uf969\u503c\u505a\u975e\u7dda\u6027\u8abf\u6574\uff0c\u4f7f\u5f97 DLG \u7279\u5fb5\u503c\u5c0d\u65bc\u5206\u985e\u8a5e\u5f59\u7684\u7279\u6027\u4e5f\u66f4\u52a0\u986f\u8457\uff0c\u4e0d\u8ad6\u76f8\u540c\u9818 \u500b \u8a5e \u5f59 \u3002 \u5de6 \u908a \u5713 \u578b \u8868 \u793a \u5019 \u9078 \u8a5e \u96c6 \u5b58 \u5728 \u65bc ICTCAS_Dic \u7684 \u8a5e \u5f59 \u96c6 \u5408 ( \u7c21 \u7a31 \u5716 3.3 \u61c9\u7528\u985e\u795e\u7d93\u7db2\u8def\u5206\u985e\u5668\u65bc\u8a5e\u5f59\u9a57\u8b49\u67b6\u69cb \u57df\u6216\u8de8\u9818\u57df\u6e2c\u8a74\uff0c\u5c0d\u65bc\u672a\u77e5\u8a5e\u5075\u6e2c\u6548\u80fd\u7684\u6539\u9032\uff0c\u5747\u6709\u986f\u8457\u7684\u6548\u679c\u3002 ICTCAS_Words) \uff0c\u7e3d\u5171\u5305\u542b 10,642 \u500b\u8a5e\u5f59\u3002\u6211\u5011\u5c07\u9019\u5169\u96c6\u5408\u7684\u4ea4\u96c6\u5171 9,802 \u500b\u8a5e\uff0c\u8996 CKIP \u65b7\u8a5e\u7cfb\u7d71 \u672c\u8ad6\u6587\u65b9\u6cd5 \u672c\u8ad6\u6587\u65b9\u6cd5 ICTCAS \u65b7\u8a5e\u7cfb\u7d71</td></tr><tr><td>\u56db\u3001\u5be6\u9a57\u5206\u6790 \u70ba \u300c\u78ba\u5b9a\u7684\u8a5e\u5f59\u300d \u3002\u5229\u7528\u9019\u4e9b\u6b63\u78ba\u7684\u8a5e\u5f59\uff0c\u5c0d\u672c\u8ad6\u6587\u7be9\u9078\u51fa\u4f86\u7684 10,000 \u500b\u8a5e\u5f59\u9032\u884c\u904e\uf984\uff0c\u7e3d (2,402) (1,486) (1,486) (1,477) \u516d\u3001\u7dd2\u8ad6</td></tr><tr><td>\u6211\u5011\u4f7f\u7528\u7b2c\u4e09\u7ae0\u7b2c\u4e00\u7bc0\u4e2d\u6240\u8ff0\u7684 CKIP_Train\u3001CKIP_Test \u53ca HUCK_Test \u8a9e\u6599\u5eab\u9032\u884c\u5be6 \u9a57\u3002\u9019\u4e09\u500b\u8a9e\u6599\u5eab\u5206\u5225\u5305\u542b\uf9ba 361,691\u3001363,382\u300154,511 \u500b\u53e5\u5b50\u3002\u7d93\u904e\u521d\u6b65\u7be9\u9078\u5019\u9078\u8a5e \u5206\u5225\u5f97\u5230 222,446\u3001224,929\u3001149160 \u500b\u5019\u9078\u8a5e\u3002\u4e26\u4e14\u5229\u7528\u500b\u5225\u8a9e\u6599\u5eab\u65b7\u8a5e\u6587\u7ae0\u4e2d\u7684\u8a5e \u5f59\uff0c\u5c0d\u65bc\u6240\u6709\u5019\u9078\u8a5e\u6a19\u8a18\u662f\u5426\u70ba\u8a5e\u5f59\uff0c\u6240\u6a19\u8a18\u7684\u8a5e\u5f59\uf969\u5206\u5225\u70ba 33,429\u300133,661\u300122,913 \u500b\u8a5e\u5f59\u3002\u95dc\u65bc\u5be6\u9a57\u8a9e\u6599\u8a73\u7d30\u8cc7\u8a0a\u5982\u8868 4.1\u3002 \u8a9e\u6599\u5eab\u7c21\u7a31 \u7528\u9014 \u53e5\uf969 \u5019\u9078\u8a5e\uf969 \u8a5e\u5f59\uf969 CKIP_Train \u5206\u985e\u5668\u8a13\u7df4 361,691 222,446 33,429 CKIP_Test \u76f8\u540c\u9818\u57df\u7684\u6e2c\u8a74 363,382 224,929 33,661 HKCU_Test \u4e0d\u540c\u9818\u57df\u7684\u6e2c\u8a74 54,511 149,160 22,913 \u8868 4.1 \u5be6\u9a57\u8a9e\u6599\u8a73\u7d30\u8cc7\u8a0a\u8868 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 1 \u5171\u6709 6,577 \u500b\u76f8\u540c\u8a5e\u5f59\uff0c\uf974\u4ee5\u300c\u78ba\u5b9a\u7b54\u6848\u300d9,802 \u500b\u8a5e\u800c\u8a00\uff0c\u672c\u8ad6\u6587\u7684\u53ec\u56de\uf961\u7d04 67.09%\u3002 \u4e94\u3001\u61c9\u7528\u65bc\u65b0\u7a4e\u9818\u57df\u7684\u672a\u77e5\u8a5e\u8403\u53d6 \u672c\u8ad6\u6587\u7814\u7a76\u7d71\u8a08\u5f0f\u7684\u8a5e\u5f59\u8403\u53d6\u65b9\u6cd5\uff0c\u5e0c\u671b\u900f\u904e\u6a5f\u5668\u5b78\u7fd2\u7684\u65b9\u6cd5\uff0c\u7d50\u5408\u4e0d\u540c\u7279\u6027\u7684\u7d71\u8a08\u7279 \u5269\u65bc\u7684 3,423 \u500b\u5019\u9078\u8a5e\uff0c\u662f\u672c\u7cfb\u7d71\u6311\u9078\u51fa\u4f86\uff0c\u4f46\u4e0d\u5728\u300c\u78ba\u5b9a\u8a5e\u5f59\u300d\u4e2d\u7684\u3002\u6211\u5011\u9032\u4e00\u6b65\u4ee5 0.9 0 0.10.20.30.40.50.60.70.80.9 1 \u7cbe \u78ba \uf961 \u53ec\u56de\uf961 No_Equ HEQ MSW \u5716 4.1 \u76f8\u540c\u9818\u57df\u4e4b\u4e0d\u540c\u6b63\u898f\u5316\u65b9\u5f0f\u7684\u6548\u80fd\u6bd4\u8f03\u5716 \u672c\u7ae0\u5c07\u63a2\u8a0e\u524d\u8ff0\u65b9\u6cd5\u61c9\u7528\u65bc\u65b0\u7a4e\u9818\u57df\u7684\u672a\u77e5\u8a5e\u8403\u53d6\u3002\u9996\u5148\uff0c\u6211\u5011\u5f9e\u7db2\u8def\u65b0\u805e\u7db2\u7ad9\u4e2d\uff0c\u4ee5 \u4eba\u5de5\u7684\u65b9\u5f0f\u9032\u884c\u8a5e\u5f59\u6a19\u8a18\uff0c\u6a19\u8a18\u51fa 1,179 \u500b\u8a5e\u5f59\uff0c\u5982\u5716 5.1(b)\u6240\u793a\u3002\u900f\u904e\u4e0a\u8ff0\u6a19\u8a18\u8a5e\u5f59 \u7684\u65b9\u6cd5\uff0c\u672c\u8ad6\u6587\u7684\u8a5e\u5f59\u8403\u53d6\u65b9\u6cd5\u7e3d\u5171\u8403\u53d6\u51fa 7,756 \u500b\u8a5e\u5f59\uff0c\u7cbe\u78ba\uf961\u7d04 77.56%\u3002 1,491 911 575 910 919 \u5fb5\uff0c\u8403\u53d6\u51fa\u672a\u77e5\u7684\u8a5e\u5f59\u3002\u9069\u6642\u66f4\u65b0\u81ea\u7136\u8a9e\u8a00\u8655\uf9e4\u7cfb\u7d71\u4e2d\u6240\u4f7f\u7528\u7684\u8a5e\u5178\uff0c\u589e\u9032\u7cfb\u7d71\u7684\u8655\uf9e4 567 \u6548\u80fd\u3002\u8a5e\u5f59\u7d44\u6210\u7684\u7d50\u69cb\u548c\u4f7f\u7528\u65b9\u5f0f\u975e\u5e38\u8907\u96dc\uff0c\u901a\u5e38\u4e26\u4e0d\u80fd\u55ae\u9760\u4e00\u7a2e\u7279\u5fb5\u503c\u4f86\u505a\u5224\u65b7\uff0c\u5f80 \u95dc\u9375\u5b57\u300c\u516b\u516b\u6c34\u707d\u300d\u641c\u5c0b 2009 \u5e74\u7684\u76f8\u95dc\u65b0\u805e\uff0c\u7576\u4f5c\u65b0\u7a4e\u9818\u57df\u7684\u6e2c\u8a74\u8a9e\u6599\u5eab(\u4ee5\u4e0b\u7c21\u7a31 \u5f80\u5fc5\u9808\u7d50\u5408\u591a\u7a2e\u7279\u5fb5\u3002\u9996\u5148\uff0c\u6211\u5011\u91dd\u5c0d\u56db\u7a2e\u4ee5\u4e0a\u7684\u7279\u5fb5\u7d44\u5408\uff0c\u9032\u884c\u5206\u6790\u5c0d\u65bc\u8a5e\u5f59\u8403\u53d6\u7684 UKW_Test) \uff0c\u7e3d\u5171 32,207 \u53e5\uff0c\u4f9d\u64da\u8a5e\u983b\u53ca\u5b57\u5143\u7d44\u9577\ufa01\u521d\u6b65\u7be9\u9078\u51fa\u5019\u9078\u8a5e\u96c6\uff0c\u7e3d\u5171\u6709 81,447 \u500b\u5019\u9078\u8a5e\u3002\u63a5\u8457\u91dd\u5c0d\u5404\u500b\u5019\u9078\u8a5e\u8a08\u7b97 DLG\u3001AV\u3001Link\u3001PreC \u56db\u7a2e\u7279\u5fb5\uf969\u503c\uff0c\u628a\u7279\u5fb5\uf969 \u503c\u6b63\u898f\u5316\u5230 0 \u81f3 1 \u4e4b\u9593\uff0c\u4e26\u4e14\u5c07 DLG \u7279\u5fb5\u503c\u9032\u884c HEQ \u6b63\u898f\u5316\uff0c\u7522\u751f\u5019\u9078\u8a5e\u7684\u7279\u5fb5\u5411 CKIP_Words (11,290) ICTCLAS_Words (10,642) \u4ea4\u96c6\u8a5e\u5f59 (9,802) \u672c\u8ad6\u6587\u65b9\u6cd5 (10,000) (a) \u6548\u80fd\u5f71\u97ff\u3002\u5be6\u9a57\u4e2d\u986f\u793a\u7576\u7d50\u5408DLG\u3001AV\u3001Link\u3001PreC\u56db\u7a2e\u7279\u5fb5\u503c\u6642\uff0c\u6703\u6709\u6700\u4f73\u7684\u5075\u6e2c (b) \u6548\u80fd\u3002\u6211\u5011\u4f7f\u7528SIGHAN2\u7af6\u8cfd\u4e2d\u7684\u8a9e\u6599\u5eab\u9032\u884c\u6e2c\u8a74\uff0c\u5c0d\u65bc\u4e2d\u7814\u9662\u8cc7\u8a0a\u6240\u8a5e\u5eab\u5c0f\u7d44\u6240\u63d0 \u5716 5.2 \u672a\u77e5\u8a5e\u5dee\u7570\u793a\u610f\u5716 \u4f9b\u7684\u8a9e\u6599\u5eab(CKIP_Test\u8a9e\u6599\u5eab)\uff0c\u5176F-Measure\u7684\uf969\u503c\u70ba60.03%\u3002 \uf97e\uff0c\u5728\u7d93\u904e CKIP_Train \u8a9e\u6599\u5eab\u8a13\u7df4\u51fa\u4f86\u7684\u8a5e\u5f59\u8403\u53d6\u6a21\u578b\uff0c\u8f38\u51fa\u4e00\u500b 0 \u81f3 1 \u7684\uf969\u503c\uff0c\u6b64 \uf969\u503c\u8868\u793a\u5019\u9078\u8a5e\u662f\u4e00\u8a5e\u5f59\u7684\u53ef\u80fd\u6027\uff0c\u4f9d\u64da\u6b64\u8f38\u51fa\uf969\u503c\u5c07\u6240\u6709\u5019\u9078\u8a5e\u9032\u884c\u6392\u5e8f\uff0c\u63a5\u8457\u7be9\u9078 \u51fa\u524d 10,000 \u500b\u5019\u9078\u8a5e\u3002 840 1,488 9,802 2,244 \u63a5\u8457\u6211\u5011\uf99c\u51fa\u53ea\u6709\u672c\u8ad6\u6587\u65b9\u6cd5\u6709\u8403\u53d6\u51fa\u4f86\uff0c\u800c\u5176\u4ed6\u5169\u5957\u7cfb\u7d71\u90fd\u6c92\u6709\u627e\u5230\u7684\u5e7e\u500b\u672a\u77e5\u8a5e\u7bc4 \u53e6\u5916\u6211 \u5011 \u91dd \u5c0d \u7d71 \u8a08 \u7279\u5fb5 \u503c \u5206 \u4f48 \u53ef \u80fd \u6703 \u6709 \u4e0d\u5339 \u914d \u7684 \u554f \u984c \uff0c \u63d0 \u51fa \uf9ba\u4f7f \u7528 \u76f4 \u65b9 \u5716 \u5747 \u5316 3,225 6,577 \u4f8b\uff0c\u5982\u8868 5.3 \u4e2d\u7684(a)\u6240\u793a\u3002\u8868 5.3(b)\u548c\u8868 5.3(c)\u5247\u5206\u5225\u662f ICTCAS \u65b7\u8a5e\u7cfb\u7d71\u8207 CKIP \u65b7\u8a5e (Histogram Equalization)\u7684\u6b63\u898f\u5316\u65b9\u6cd5\uff0c\u4f7f\u5f97\u6e2c\u8a74\u8207\u8a13\u7df4\u7279\u5fb5\u503c\u5206\u4f48\u80fd\u4e92\u76f8\u5339\u914d\uff0c\u89e3 1,179 \u7cfb\u7d71\u6709\u627e\u5230\u7684\u672a\u77e5\u8a5e\uff0c\u800c\u672c\u8ad6\u6587\u65b9\u6cd5\u4e26\u672a\u627e\u5230\u7684\u672a\u77e5\u8a5e\u7bc4\u4f8b\u3002\u5f9e\u8868 5.3(a)\u53ef\u4ee5\u770b\u51fa\uff0c\u6709 \u6c7a\u8a9e\u6599\u5eab\u5927\u5c0f\u6216\u9818\u57df\u4e0d\u540c\u6240\u9020\u6210\u7279\u5fb5\u503c\u7bc4\u570d\u8b8a\u52d5\u53ca\u5206\u4f48\u5dee\u7570\u7684\u554f\u984c\u3002\u4e0d\u5fc5\u56e0\u70ba\u9818\u57df\u7684\u5dee (\u4e00)\u3001\u6a19\u8a18\u8a5e\u5f59\u7684\u65b9\u6cd5 \u8981\u5c0d\u65b0\u7a4e\u9818\u57df\u5206\u6790\u8a5e\u5f59\u8403\u53d6\u6548\u80fd\uff0c\u6700\u5927\u7684\u56f0\u96e3\u662f\u5728\u4e2d\u6587\u4e0a\u8a5e\u5f59\u4e26\u6c92\u6709\u5171\u901a\u7684\u5b9a\u7fa9\uff0c\u4e0d\u540c (a) \u5f88\u591a\u65b0\u7522\u751f\u51fa\u4f86\u7684\u8a5e\u5f59\uff0c\u4f8b\u5982\u3006 \u300c\u881f\u7b46\u5c0f\u65b0\u300d \u3001 \u300c\u91d1\u878d\u6d77\u562f\u300d\u7b49\u5c08\u6709\u540d\u8a5e\uff0c\u5176\u5be6\u5f88\u96e3\u900f\u904e \u7570\u800c\u91cd\u65b0\u8a13\u7df4\u8a5e\u5f59\u8403\u53d6\u6a21\u578b\uff0c\u964d\u4f4e\u5c0d\u65bc\u7279\u5b9a\u8a13\u7df4\u8a9e\u6599\u5eab\u7684\u4f9d\u8cf4\u6027\u3002\u4f7f\u5f97\u672c\u8ad6\u6587\u7684\u8a5e\u5f59\u8403 (b) \u8a9e\u6cd5\u7684\u898f\u5247\u62bd\u53d6\u51fa\u4f86\uff0c\u4f46\u5b83\u5011\u5177\u6709\u660e\u986f\u7684\u7d71\u8a08\u7279\u6027\uff0c\u53ef\u4ee5\u5229\u7528\u672c\u8ad6\u6587\u7684\u7d71\u8a08\u7279\u5fb5\u62bd\u53d6\u51fa \u53d6\u65b9\u6cd5\u66f4\u5177\u4e00\u822c\u6027\uff0c\u80fd\u91dd\u5c0d\u5404\u7a2e\u4e0d\u540c\u9818\u57df\uff0c\u8403\u53d6\u51fa\u5404\u500b\u9818\u57df\u5e38\u4f7f\u7528\u7684\u8a5e\u5f59\u3002\u5c0d\u65bc\u4e2d\u7814\u9662</td></tr><tr><td>\u9996\u5148\uff0c\u6211\u5011\u5c0d\u65bc\u76f8\u540c\u9818\u57df\u8a9e\u6599\u5eab\uff0c\u9032\u884c\u56db\u7a2e\u4ee5\u4e0a\u7684\u7279\u5fb5\u7d44\u5408\u5be6\u9a57\u3002\u4ee5 CKIP_Train \u8a13\u7df4 \u7684\u65b7\u8a5e\u7cfb\u7d71\u6216\u8a5e\u5178\u6240\u4f7f\u7528\u7684\u8a5e\u5f59\u5b9a\u7fa9\u53ef\u80fd\u6703\u6709\u6240\u51fa\u5165\uff0c\u4f8b\u5982\u8868 5.1 \u6240\u793a\uff0c\u6b64\u8868\uf99c\u51fa\u90e8\u5206 \u5716 5.1 \u65b0\u7a4e\u9818\u57df\u4e4b\u8a5e\u5f59\u6a19\u8a18\u65b9\u6cd5\u793a\u610f\u5716 \u4f86\u3002CKIP \u548c ICTCAS \u65b7\u8a5e\u7cfb\u7d71\u7d50\u5408\uf9ba\u7d71\u8a08\u548c\u8a9e\u7fa9\u7684\u7279\u6027\u5075\u6e2c\u672a\u77e5\u8a5e\uff0c\u4e5f\u672a\u80fd\u62bd\u51fa\u9019\u4e9b \u8cc7\u8a0a\u6240\u8a5e\u5eab\u5c0f\u7d44\u53ca\u9999\u6e2f\u57ce\u5e02\u5927\u5b78\u6240\u63d0\u4f9b\u7684\u8a9e\u6599\u5eab(CKIP_Test\u8a9e\u6599\u5eab\u3001HCKU_Test\u8a9e\u6599</td></tr><tr><td>\u5206\u985e\u5668\uff0c\u4e26\u4ee5 CKIP_Test \u9032\u884c\u6e2c\u8a74\u3002\u5be6\u9a57\u7d50\u679c\u5982\u8868 4.2\u3002\u5728\u8868 4.2 \u4e2d\uff0cALL \u4ee3\u8868\u4f7f\u7528\uf9ba \u4e94\u7a2e\u7279\u5fb5\u503c\uff0c\u5176\u4ed6\u5247\u5206\u5225\u4ee3\u8868\u522a\u9664\u5176\u4e2d\u4e00\u7a2e\u7279\u5fb5\uff0c\u4f8b\u5982\u3006No_LogC \u8868\u793a\u522a\u9664 LogC \u7279 CKIP \u8207 HKCU \u4e0d\u540c\u8a5e\u5f59\u5b9a\u7fa9\u7684\u4f8b\u5b50\uff0c \u300c\u8a5e\u5f59\u300d\u6b04\u4f4d\u8868\u793a\u5c0d\u6b64\u5b57\u7684\u7d44\u7684\u8a5e\u5f59\u5b9a\u7fa9\uff0c\u5982\u5728 CKIP \u6b04\u4e2d\u300c\u5976\u7c89 \u9322\u300d\u8868\u793a\u53e5\u5b50\u4e2d\u7684\u300c\u5976\u7c89\u9322\u300d\u662f\u7531\u5169\u500b\u8a5e\u5f59\u300c\u5976\u7c89\u300d\u8207\u300c\u9322\u300d\u7d44\u6210\u7684\u3005 \u5177\u7684\u6709\u7d71\u8a08\u7279\u5fb5\u7684\u672a\u77e5\u8a5e\uff0c\u9019\u986f\u793a\uf9ba\u672c\u8ad6\u6587\u7d50\u5408\u591a\u7a2e\u4e0d\u540c\u7279\u6027\u7684\u7d71\u8a08\u7279\u5fb5\uff0c\u7684\u78ba\u53ef\u4ee5\u66f4 \u5eab)\uff0cF-Measure\u5206\u5225\u53ef\u4ee5\u9054\u523068.43\uff05\u548c71.40%\u3002\u540c\u6642\u6211\u5011\u4e5f\u767c\u73fe\uff0c\u91dd\u5c0dDLG\u505a\u76f4\u65b9\u5716 (\u4e8c)\u3001\u65b0\u7a4e\u9818\u57df\u4e4b\u672a\u77e5\u8a5e\u8403\u53d6\u5206\u6790 \u53ef\u9760\u5730\u8403\u53d6\u51fa\u4e00\u4e9b\u9808\u4f9d\u8cf4\u7d71\u8a08\u7279\u5fb5\u7684\u65b0\u7a4e\u8a5e\u5f59\u3002\u4e0d\u904e\u9019\u5169\u5957\u7cfb\u7d71\u5728\u7279\u6b8a\u985e\u5225\u7684\u672a\u77e5\u8a5e\u62bd \u5747\u5316\uff0c\u4e0d\u8ad6\u662f\u5728\u76f8\u540c\u9818\u57df\u6216\u8de8\u9818\u57df\u7684\u6e2c\u8a74\uff0c\u5747\u53ef\u4ee5\u6539\u9032\u8a5e\u5f59\u8403\u53d6\u7684\u6548\u80fd\u3002</td></tr><tr><td>\u91dd\u5c0d\u7d50\u5408 DLG\u3001AV\u3001Link\u3001PreC\u3002 \u7570\uff0c\u4f7f\u5f97\u8a31\u591a\u5019\u9078\u8a5e\u7684\u6c7a\u7b56\uf969\u503c y (\u5716 3.3 \u4e2d\u5206\u985e\u5668\u7522\u751f\u7684\u6c7a\u7b56\uf969\u503c y) \u504f\u5c0f\u4e14\u7686\u76f8\u540c\uff0c\u7121 \u6cd5\u53ef\u9760\u5206\u985e\u7684\u554f\u984c\u3002\u6240\u4ee5\u7d93\u904e\u6b63\u898f\u5316\u5f8c\uff0c\u53ef\u4ee5\u6709\u6548\u7684\u8b58\u5225\u9019\u4e9b\u5019\u9078\u8a5e\u3002\u4e26\u4e14\u5f9e\u5716\u4e2d\u53ef \u4ee5\u660e\u986f\u7684\u770b\u51fa\u7d93\u904e\u975e\u7dda\u6027 HEQ \u8f49\u63db\u5f8c\uff0c\u4f7f\u5f97\u6548\u80fd\u5927\u5e45\u7684\u63d0\u5347\uff0c\u6700\u4f73\u7684 F-Measure \u53ef\u9054 71.40%\u3002 \u53cd\u4e4b\uff0c\u5728 HKCU \u6b04\u4e2d\uff0c\u5247\u8a8d\u70ba\u300c\u5976\u7c89\u9322\u300d\u61c9\u8a72\u88ab\u8996\u70ba\u55ae\u4e00\u500b\u8a5e\u5f59\u3002\u56e0\u6b64\uff0c\u6211\u5011\u7684\u7b56\u7565 \u662f\u4f7f\u7528\u5169\u500b\u7dda\u4e0a\u65b7\u8a5e\u7cfb\u7d71\u6240\u5171\u540c\u7522\u751f\u7684\u8a5e\u5f59\u4f5c\u70ba\u300c\u78ba\u5b9a\u7684\u7b54\u6848\u300d \uff0c\u518d\u8f14\u4ee5\u4eba\u5de5\u6a19\u8a18\u3002\u9996 \u5148\uff0c\u6211\u5011\u5c07 UKW_Test\uff0c\u5206\u5225\u900f\u904e\u5169\u5957\u5177\u6709\u5075\u6e2c\u65b0\u8a5e\u80fd\u529b\u7684\u7e41\u9ad4\u65b7\u8a5e\u7cfb\u7d71\u9032\u884c\u65b7\u8a5e\u3002\u9019 \u5169\u5957\u7cfb\u7d71\u70ba CKIP \u8207\u4e2d\u570b\u79d1\u5b78\u9662\u8a08\u7b97\u6280\u8853\u6240(Institute of Computing Technology Chinsee Academy of Science, ICTCAS)\u6240\u63d0\u4f9b\u7684\u65b7\u8a5e\u670d\u52d9\u3002\u4e4b\u5f8c\uff0c\u5c07\u65b7\u8a5e\u5f8c\u7522\u751f\u7684\u8a5e\u5f59\uff0c\u8996\u70ba \u63a5\u4e0b\u4f86\u6211\u5011\u5c0d\u672c\u8ad6\u6587\u8a5e\u5f59\u8403\u53d6\u65b9\u6cd5\u8207\u5169\u5957\u65b7\u8a5e\u7cfb\u7d71\u6240\u8403\u53d6\u51fa\u4f86\u7684\u672a\u77e5\u8a5e\uff0c\u9032\u884c\u5206\u6790\u6bd4\u8f03\u3002 \u672c\u8ad6\u6587\u5b9a\u7fa9\u7684\u672a\u77e5\u8a5e\u5373\u70ba\u8a13\u7df4\u8a9e\u6599\u5eab(CKIP_Train \u8a9e\u6599)\u4e2d\u672a\u51fa\u73fe\u7684\u8a5e\u5f59\u3002\u56e0\u6b64\u6211\u5011 \u5c07\u5404\u65b9\u6cd5\u8403\u53d6\u51fa\u4f86\u7684\u8a5e\u5f59\u8207\u8a13\u7df4\u8a9e\u6599\u5eab\u7684\u8a5e\u5f59\u505a\u6bd4\u8f03\uff0c\u522a\u53bb\u8a13\u7df4\u8a9e\u6599\u5eab\u4e2d\u5df2\u51fa\u73fe\u7684\u8a5e \u53d6\u6548\u80fd\u6bd4\u672c\u8ad6\u6587\u7684\u65b9\u6cd5\u4f73\uff0c\u4f8b\u5982\u3006\u4eba\u540d\u3001\u5730\u540d\u6216\u662f\u5177\u6709\u7279\u6b8a\u6587\u6cd5\u7d50\u69cb\u7684\u8a5e\u5f59\uff0c\u5982\u300c\u8607\u7e23 \u6700\u5f8c\u6211\u5011\u5c07\u8a5e\u5f59\u8403\u53d6\u7684\u65b9\u6cd5\u61c9\u65bc\u7528\u8403\u53d6\u65b0\u7a4e\u9818\u57df\u7684\u672a\u77e5\u8a5e\uff0c\u4e26\u8207\u4e2d\u592e\u7814\u7a76\u9662\u8cc7\u8a0a\u6240\u8a5e\u5eab \u9577\u300d \u3001 \u300c\u7d93\u767c\u5c40\u300d\u7b49\u3002\u9019\u662f\u7531\u65bc\u672c\u8ad6\u6587\u662f\u7d14\u7cb9\u4ee5\u7d71\u8a08\u7279\u5fb5\u70ba\u4e3b\u8981\u8403\u53d6\u65b9\u6cd5\uff0c\u4e26\u672a\u4f7f\u7528\u4efb\u4f55 \u5c0f\u7d44\u548c\u4e2d\u570b\u79d1\u5b78\u9662\u8a08\u7b97\u6280\u8853\u6240\u63d0\u4f9b\u7684\u65b7\u8a5e\u7cfb\u7d71\u62bd\u53d6\u7684\u672a\u77e5\u8a5e\u9032\u884c\u5206\u6790\u6bd4\u8f03\uff0c\u6211\u5011\u767c\u73fe\u672c \u6587\u6cd5\u898f\u5247\u3002\u9019\u6a23\u7684\u65b9\u6cd5\u61c9\u8a72\u548c\u4ee5\u8a9e\u6cd5\u898f\u5247\u62bd\u8a5e\u7684\u67b6\u69cb\u7522\u751f\u5f88\u5927\u7684\u4e92\u88dc\u6027\u3002 \u8ad6\u6587\u65b9\u6cd5\u8207\u5176\u5169\u5957\u65b7\u8a5e\u7cfb\u7d71\u5177\u6709\u4e92\u88dc\u7684\u7279\u6027\uff0c\u53ef\u4ee5\u8403\u53d6\u51fa\u5177\u6709\u5f37\u70c8\u7684\u7d71\u8a08\u8a5e\u5f59\u7279\u6027\u4e14\u96e3 \u5f59\uff0c\u4f5c\u70ba\u5404\u500b\u65b9\u6cd5\u8403\u53d6\u51fa\u7684\u672a\u77e5\u8a5e\u96c6\u5408\u3002\u5982\u8868 5.2 \u6240\u793a\u3002\u672c\u8ad6\u6587\u8a5e\u5f59\u8403\u53d6\u65b9\u6cd5\u7be9\u9078\u51fa 1,486 \u500b\u672a\u77e5\u8a5e\uff0cCKIP \u65b7\u8a5e\u7cfb\u7d71\u7be9\u9078\u51fa 2,402 \u500b\u672a\u77e5\u8a5e\uff0c\u800c ICTCAS \u65b7\u8a5e\u7cfb\u7d71\u7be9\u9078\u51fa 1,477 \u4ee5\u900f\u904e\u8a9e\u610f\u7684\u65b9\u5f0f\u8403\u53d6\u51fa\u4f86\u7684\u672a\u77e5\u8a5e\uff0c\u4f8b\u5982\u3006\u300c\u6d77\u89d27\u865f\u300d\u3001\u300c\u91d1\u878d\u6d77\u562f\u300d\u7b49\u672a\u77e5\u8a5e\u3002</td></tr></table>", | |
| "html": null, | |
| "text": "\u3001Link\u3001PreC\u3001DLG\uff0c\u7d93\u516c\u5f0f(2.6)\u6b63\u898f\u5316\uff0c\u4e26\u4e14\u5c07 DLG \u9032\u4e00\u6b65\u505a HEQ \u6216 MSW \u6b63\u898f\u5316\uff0c\u7d93\u7279\u5fb5\u9078\u53d6\u5f8c\u4f5c\u70ba\u6b64\u5206\u985e\u5668\u7684\u8f38\u5165\u7279\u5fb5\u3002\u8f38\u5165\u7279\u5fb5 x \u63a5\u8457\uff0c\u6211\u5011\u5c0d\u8de8\u9818\u57df\u8a9e\u6599\u5eab\u9032\u884c\u5be6\u9a57\uff0c\u4ee5 CKIP_Train \u8a13\u7df4\u5206\u985e\u5668\uff0cHKCU_Test \u9032\u884c\u6e2c \u8a74\u3002\u5be6\u9a57\u7d50\u679c\u5982\u5716 4.2\u3002\u5716\u4e2d\u7684\u9ed1\u8272\u5be6\u7dda\u66f2\u7dda No_Equ \u662f\u6c92\u505a\u6b63\u898f\u5316\u7684\u6548\u80fd\u66f2\u7dda\u3001\u7070\u8272 \u5be6\u7dda\u66f2\u7dda HEQ \u8868\u793a\u91dd\u5c0d DLG \u7279\u5fb5\u503c\u505a\u5b8c HEQ \u6b63\u898f\u5316\u5f8c\u7684\u6548\u80fd\u66f2\u7dda\uff0c\u7070\u8272\u865b\u7dda\u66f2\u7dda MSW \u8868\u793a\u91dd\u5c0d DLG \u7279\u5fb5\u503c\u505a\u5b8c MSW \u6b63\u898f\u5316\u5f8c\u7684\u6548\u80fd\u66f2\u7dda\u3002\u5f9e\u5716\u4e2d\u53ef\u4ee5\u770b\u51fa\u7576 Recall \u503c\u8f03\u9ad8\u6642(\u9580\u6abb\u503c \u03b7 \u8f03\u4f4e\u6642)\uff0c\u7d93\u904e MSW \u8f49\u63db\u7684\u6548\u80fd\u63d0\u5347\u4e00\u4e9b\uff0c\u9019\u662f\u56e0\u70ba\u7d71\u8a08\u5206\u4f48\u5dee", | |
| "num": null | |
| } | |
| } | |
| } | |
| } |