| { | |
| "paper_id": "2020", | |
| "header": { | |
| "generated_with": "S2ORC 1.0.0", | |
| "date_generated": "2023-01-19T07:27:24.410376Z" | |
| }, | |
| "title": "Chinese Healthcare Named Entity Recognition Based on Graph Neural Networks \u76e7\u6bc5 \uf02a \u3001\u674e\u9f8d\u8c6a \uf02a", | |
| "authors": [ | |
| { | |
| "first": "Yi", | |
| "middle": [], | |
| "last": "Lu", | |
| "suffix": "", | |
| "affiliation": { | |
| "laboratory": "", | |
| "institution": "National Central University", | |
| "location": {} | |
| }, | |
| "email": "" | |
| }, | |
| { | |
| "first": "Lung-Hao", | |
| "middle": [], | |
| "last": "Lee", | |
| "suffix": "", | |
| "affiliation": { | |
| "laboratory": "", | |
| "institution": "National Central University", | |
| "location": {} | |
| }, | |
| "email": "[email protected]" | |
| }, | |
| { | |
| "first": "\uf02a", | |
| "middle": [], | |
| "last": "\u570b\u7acb\u4e2d\u592e\u5927\u5b78\u96fb\u6a5f\u5de5\u7a0b\u7814\u7a76\u6240", | |
| "suffix": "", | |
| "affiliation": { | |
| "laboratory": "", | |
| "institution": "National Central University", | |
| "location": {} | |
| }, | |
| "email": "" | |
| } | |
| ], | |
| "year": "", | |
| "venue": null, | |
| "identifiers": {}, | |
| "abstract": "Named Entity Recognition (NER) focuses on locating the mentions of name entities and classifying their types, usually referring to proper nouns such as persons, places, organizations, dates, and times. The NER results can be used as the basis for relationship extraction, event detection and tracking, knowledge graph building, and question answering system. NER studies usually regard this research topic as a sequence labeling problem and learns the labeling model through the large-scale corpus. We propose a GGSNN (Gated Graph Sequence Neural Networks) model for Chinese healthcare NER. We derive a character representation based on multiple embeddings in different granularities from the radical, character to word levels. An adapted gated graph sequence neural network is involved to incorporate named entity information in the dictionaries. A standard BiLSTM-CRF is then used to identify named entities and classify their types in the healthcare domain. We firstly crawled articles from websites that provide healthcare information, online health-related news and medical question/answer forums. We then randomly selected partial sentences to retain content diversity. It includes 30,692 sentences with a total of around 1.5 million characters or 91.7 thousand words. After manual annotation, we have 68,460 named entities across 10 entity types: body, symptom, instrument, examination, chemical, disease, drug, supplement, treatment, and time. Based on further experiments and error analysis, our proposed method achieved the best F1-score of 75.69% that outperforms previous models including the BiLSTM-CRF, Lattice, Gazetteers, and ME-CNER. In summary, our GGSNN model is an effective and efficient solution for the Chinese healthcare NER task.", | |
| "pdf_parse": { | |
| "paper_id": "2020", | |
| "_pdf_hash": "", | |
| "abstract": [ | |
| { | |
| "text": "Named Entity Recognition (NER) focuses on locating the mentions of name entities and classifying their types, usually referring to proper nouns such as persons, places, organizations, dates, and times. The NER results can be used as the basis for relationship extraction, event detection and tracking, knowledge graph building, and question answering system. NER studies usually regard this research topic as a sequence labeling problem and learns the labeling model through the large-scale corpus. We propose a GGSNN (Gated Graph Sequence Neural Networks) model for Chinese healthcare NER. We derive a character representation based on multiple embeddings in different granularities from the radical, character to word levels. An adapted gated graph sequence neural network is involved to incorporate named entity information in the dictionaries. A standard BiLSTM-CRF is then used to identify named entities and classify their types in the healthcare domain. We firstly crawled articles from websites that provide healthcare information, online health-related news and medical question/answer forums. We then randomly selected partial sentences to retain content diversity. It includes 30,692 sentences with a total of around 1.5 million characters or 91.7 thousand words. After manual annotation, we have 68,460 named entities across 10 entity types: body, symptom, instrument, examination, chemical, disease, drug, supplement, treatment, and time. Based on further experiments and error analysis, our proposed method achieved the best F1-score of 75.69% that outperforms previous models including the BiLSTM-CRF, Lattice, Gazetteers, and ME-CNER. In summary, our GGSNN model is an effective and efficient solution for the Chinese healthcare NER task.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Abstract", | |
| "sec_num": null | |
| } | |
| ], | |
| "body_text": [ | |
| { | |
| "text": "\u547d\u540d\u5be6\u9ad4\u8fa8\u8b58", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "\u7dd2\u8ad6 (Introduction)", | |
| "sec_num": "1." | |
| }, | |
| { | |
| "text": ") \u3001 \u6700 \u5927 \u5316 \u71b5 \u99ac \u53ef \u592b \u6a21 \u578b (Maximum Entropy Markov Model, MEMM) (Toutanova & Manning, 2000) \u548c\u689d\u4ef6\u96a8\u6a5f\u5834\u57df(Conditional Random Field, CRF) (Lafferty, McCallum & Pereira, 2001 ", | |
| "cite_spans": [ | |
| { | |
| "start": 59, | |
| "end": 86, | |
| "text": "(Toutanova & Manning, 2000)", | |
| "ref_id": "BIBREF12" | |
| }, | |
| { | |
| "start": 126, | |
| "end": 161, | |
| "text": "(Lafferty, McCallum & Pereira, 2001", | |
| "ref_id": "BIBREF4" | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "\u7dd2\u8ad6 (Introduction)", | |
| "sec_num": "1." | |
| }, | |
| { | |
| "text": "EQUATION", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [ | |
| { | |
| "start": 0, | |
| "end": 8, | |
| "text": "EQUATION", | |
| "ref_id": "EQREF", | |
| "raw_str": "\u672c\u7814\u7a76\u63d0\u51fa\u7684\u9580\u63a7\u5716\u5e8f\u5217\u795e\u7d93\u7db2\u8def(GGSNN)\u6a21\u578b\u67b6\u69cb\u5982\u4e0b\u5716 1\uff0c\u6b64\u6a21\u578b\u4f7f\u7528\u4e86\u76ee\u524d\u4e3b\u6d41 \u7684 BiLSTM-CRF \u4f5c\u70ba\u6a21\u578b\u7684\u57fa\u790e\u67b6\u69cb\uff0c\u4e26\u5c0d\u5176\u505a\u5ef6\u4f38\uff0c\u6a21\u578b\u7e3d\u5171\u5206\u70ba\u56db\u5c64\u3002 \u5716 1. GGSNN \u6a21\u578b\u67b6\u69cb [Figure 1. GGSNN model architecture] 3.1 \u591a\u91cd\u5d4c\u5165\u5c64 (Multiple Embeddings Layer) \u900f\u904e\u7d44\u5408\u5b57\u5d4c\u5165\u3001\u8a5e\u5d4c\u5165\u4ee5\u53ca\u90e8\u9996\u5d4c\u5165\u5f62\u6210\u591a\u91cd\u5d4c\u5165\uff0c\u5176\u4e2d\u5b57\u5d4c\u5165\u3001\u90e8\u9996 \u5d4c\u5165\u4ee5\u53ca\u8a5e\u5d4c\u5165\u7684\u8655\u7406\u5206\u5225\u5982\u4e0b\u5217\u6558\u8ff0\uff0c\u5047\u8a2d\u8f38\u5165\u53e5\u5b50\u5b57\u6578\u9577\u5ea6\u70ba n\uff1a (1)\u3001\u5b57\u5d4c\u5165 (Character Embedding)\uff1a \u8f38\u5165\u5b57\u5e8f\u5217 , , , \u2026 , \uff0c\u5206\u5225\u7d93\u904e BiLSTM \u4ee5\u53ca\u5377\u7a4d\u904b\u7b97\u5f8c\uff0c\u5c07\u5169\u8005\u7d44\u6210\u65b0\u7684 \u5b57\u5d4c\u5165\u7279\u5fb5\u5e8f\u5217\uff0c\u5f97\u5230\u5e8f\u5217 , , , \u2026 , \uff0c\u7531\u65bc\u6bcf\u500b\u5b57\u53ef\u80fd\u8207\u9577\u8ddd\u96e2\u7684\u53e6\u4e00\u500b\u5b57 \u6216\u662f\u9644\u8fd1\u7684\u5b57\u6709\u6240\u95dc\u806f\uff0c\u56e0\u6b64\u900f\u904e BiLSTM \u53ef\u4ee5\u6355\u6349\u5230\u9577\u8ddd\u96e2\u7684\u8cc7\u8a0a\uff0c\u800c\u5377\u7a4d\u904b\u7b97\u53ef\u4ee5 \u6355\u6349\u5230\u77ed\u8ddd\u96e2\u7684\u8cc7\u8a0a\u3002 \u76e7\u6bc5\u8207\u674e\u9f8d\u8c6a , , , \u2026 , = (1) , , , \u2026 , = (2) = \u2295 (3) (2)\u3001\u90e8\u9996\u5d4c\u5165 (Radical Embedding)\uff1a \u8f38 \u5165 \u90e8 \u9996 \u5e8f \u5217 , , , \u2026 , \uff0c \u7d93 \u904e \u5377 \u7a4d \u904b \u7b97 \u5f8c \uff0c \u5f97 \u5230 \u65b0 \u7684 \u90e8 \u9996 \u7279 \u5fb5 \u5e8f \u5217 , , , \u2026 , \uff0c\u7531\u65bc\u6bcf\u500b\u90e8\u9996\u591a\u534a\u8207\u9644\u8fd1\u7684\u5b57\u6709\u95dc\uff0c\u56e0\u6b64\u5377\u7a4d\u904b\u7b97\u53ef\u4ee5\u6355\u6349\u5230\u77ed\u8ddd\u96e2\u7684 \u8cc7\u8a0a\u3002 , , , \u2026 , = (4) (3)\u3001\u8a5e\u5d4c\u5165 (Word Embedding)\uff1a \u7531\u65bc\u6a21\u578b\u662f\u4ee5\u5b57\u70ba\u57fa\u790e\u4f5c\u70ba\u8f38\u5165\uff0c\u800c\u540c\u4e00\u500b\u5b57\u7d44\u6210\u7684\u4e0d\u540c\u8a5e\u8a9e\u53ef\u80fd\u6709\u4e0d\u540c\u7684\u610f\u601d\uff0c\u56e0\u6b64 \u76f8\u540c\u5b57\u7684\u8cc7\u8a0a\uff0c\u52a0\u5165\u4e0d\u540c\u8a5e\u7684\u8cc7\u8a0a\uff0c\u53ef\u4ee5\u89e3\u6c7a\u6b64\u7a2e\u60c5\u6cc1\uff0c\u800c\u8a5e\u7684\u8cc7\u8a0a\u662f\u5c6c\u65bc\u8f03\u9ad8\u968e\u7684\u7279 \u5fb5\uff0c\u56e0\u6b64\u672c\u7814\u7a76\u76f4\u63a5\u5c07\u5176\u4f5c\u5408\u4f75\uff0c\u4e0d\u505a\u984d\u5916\u7684\u8655\u7406\u3002 W = , , , \u2026 ,", | |
| "eq_num": "(5)" | |
| } | |
| ], | |
| "section": "\u7dd2\u8ad6 (Introduction)", | |
| "sec_num": "1." | |
| }, | |
| { | |
| "text": "EQUATION", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [ | |
| { | |
| "start": 0, | |
| "end": 8, | |
| "text": "EQUATION", | |
| "ref_id": "EQREF", | |
| "raw_str": "\u6700\u7d42\u5c07\u5b57\u7279\u5fb5\u5e8f\u5217\u3001\u90e8\u9996\u7279\u5fb5\u5e8f\u5217\u4ee5\u53ca\u8a5e\u7279\u5fb5\u5e8f\u5217\uff0c\u7d44\u5408\u6210\u591a\u91cd\u5d4c\u5165\u5982\u4e0b\uff1a = \u2295 \u2295", | |
| "eq_num": "(6)" | |
| } | |
| ], | |
| "section": "\u7dd2\u8ad6 (Introduction)", | |
| "sec_num": "1." | |
| }, | |
| { | |
| "text": "\u5176\u4e2d \u4ee3\u8868\u7d93\u904e\u8655\u7406\u5f8c\u7684\u5b57\u5d4c\u5165\uff0c \u4ee3\u8868\u7d93\u904e\u8655\u7406\u5f8c\u7684\u90e8\u9996\u5d4c\u5165\uff0c \u4ee3\u8868\u8a5e\u5d4c\u5165\uff0c \u4ee3 \u8868\u62fc\u63a5\u5f8c\u7684\u591a\u91cd\u5d4c\u5165\u3002 ", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "\u7dd2\u8ad6 (Introduction)", | |
| "sec_num": "1." | |
| }, | |
| { | |
| "text": "\u5728\u672c\u7814\u7a76\u4e2d\u63a1\u7528\u6539\u826f\u5f0f GGSNN \u5b78\u7fd2\u53e5\u5b50\u5716\u7d50\u69cb\u5316\u5f8c\u7684\u8a0a\u606f\uff0c\u8207 Li \u7b49\u4eba(2016) \u6240\u63d0\u51fa\u7684 GGSNN \u4e0d\u540c\u4e4b\u8655\u5728\u65bc\u6539\u826f\u5f0f\u7684 GGSNN \u53ef\u4ee5\u7d66\u4e88\u908a\u4e0a\u6a19\u7c64\u4e0d\u540c\u7684\u6b0a\u91cd\uff0c\u900f\u904e\u6539\u826f\u5f0f\u7684 GGSNN \u53ef\u4ee5\u5c07\u52a0\u5165\u591a\u500b\u5b57\u5178\u7684\u8a0a\u606f\uff0c\u4e26\u4e14\u7d66\u4e88\u4e0d\u540c\u7684\u5b57\u5178\u4e0d\u540c\u7684\u6b0a\u91cd\u3002\u4f46\u7531\u65bc\u786c\u9ad4\u7684 \u9650\u5236\uff0c\u6211\u5011\u7121\u6cd5\u4e0d\u53d7\u9650\u5236\u7684\u8ffd\u52a0\u591a\u500b\u5b57\u5178\uff0c\u56e0\u6b64\u8207 Ding \u7b49\u4eba(2019) \u7684\u5b57\u5178\u7de8\u6392\u65b9\u5f0f\u4e0d \u540c\uff0c\u672c\u7814\u7a76\u5c07\u5b57\u5178\u88e1\u7684\u8a5e\u5f59\u4f9d\u7167\u5b57\u6578\u505a\u5206\u985e\uff0c\u7e3d\u5171\u5206\u6210\u4e94\u500b\u5b57\u5178\u3002 \u5728\u9019\u5c64\u7d50\u69cb\u4e2d\u9996\u5148\u6703\u5229\u7528\u5b57\u5178\uff0c\u900f\u904e\u5b57\u4e32\u6bd4\u5c0d\u7522\u751f\u591a\u7dad\u6709\u5411\u5716\uff0c\u5efa\u69cb\u51fa\u7684\u591a\u7dad\u6709\u5411 \u5716 (Multi-digraph) \u7bc4\u4f8b\u5982\u5716 2\u3002\u7d66\u5b9a\u4e00\u500b\u591a\u7dad\u6709\u5411\u5716 \u2236 , ,", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "\u9580\u63a7\u5716\u5e8f\u5217\u795e\u7d93\u7db2\u8def\u5c64 (GGSNN Layer)", | |
| "sec_num": "3.2" | |
| }, | |
| { | |
| "text": "EQUATION", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [ | |
| { | |
| "start": 0, | |
| "end": 8, | |
| "text": "EQUATION", | |
| "ref_id": "EQREF", | |
| "raw_str": "\u7531\u8f38\u5165\u53e5\u5b50\u7684\u539f\u59cb\u5b57\u5e8f\u5217\u8a0a\u606f\u53ef\u4ee5\u5f97\u5230\u76f8\u9130\u77e9\u9663 \uff0c\u800c\u7531\u4e0d\u540c\u7684\u5b57\u5178\u53ef\u4ee5\u5f97\u5230\u5176\u76f8 \u5c0d \u61c9 \u7684 \u76f8 \u9130 \u77e9 \u9663 \uff0c \u4f9d \u7167 \u672c \u7814 \u7a76 \u7684 \u5b57 \u5178 \u5206 \u985e \u65b9 \u5f0f \u53ef \u4ee5 \u5f97 \u5230 \u76f8 \u9130 \u77e9 \u9663 \u3001 \u3001 \u3001 \u4ee5\u53ca \uff0c\u5176\u4e2d \u4ee3\u8868\u7684\u70ba\u5b57\u5178\u8a5e\u5f59\u5b57\u6578\u9577\u5ea6\u70ba 1 \u7684\u76f8\u9130\u77e9\u9663\uff0c\u5176\u9918\u4f9d\u6b64\u985e \u63a8\u3002 \u5728\u672c\u7814\u7a76\u4e2d\uff0c\u4e0d\u540c\u5b57\u5178\u7684\u76f8\u9130\u77e9\u9663\u6703\u5206\u5225\u7d66\u5b9a\u4e0d\u540c\u7684\u6b0a\u91cd\uff0c\u6b0a\u91cd\u7531\u4ee5\u4e0b\u7684\u516c\u5f0f\u6c7a \u5b9a\uff1a , , , , , = \u03c3 , , , , ,", | |
| "eq_num": "(7)" | |
| } | |
| ], | |
| "section": "\u9580\u63a7\u5716\u5e8f\u5217\u795e\u7d93\u7db2\u8def\u5c64 (GGSNN Layer)", | |
| "sec_num": "3.2" | |
| }, | |
| { | |
| "text": "\u5176\u4e2d , , , , ,", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "\u9580\u63a7\u5716\u5e8f\u5217\u795e\u7d93\u7db2\u8def\u5c64 (GGSNN Layer)", | |
| "sec_num": "3.2" | |
| }, | |
| { | |
| "text": "EQUATION", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [ | |
| { | |
| "start": 0, | |
| "end": 8, | |
| "text": "EQUATION", | |
| "ref_id": "EQREF", | |
| "raw_str": "\u70ba\u53ef\u4ee5\u88ab\u8a13\u7df4\u7684\u53c3\u6578\uff0c\u4e26\u4e14\u900f\u904e sigmod \u51fd\u6578\u4f7f\u5176\u8f49\u63db \u6210\u6700\u5f8c\u7684\u6b0a\u91cd , ,", | |
| "eq_num": ", , , \uff0c\u5c07\u4e0d\u540c\u7684\u5168\u6b0a\u5206\u5225\u4e58\u4e0a\u76f8\u5c0d\u61c9\u7684\u76f8\u9130\u77e9\u9663\uff0c\u5373 \u53ef\u7372\u5f97\u6700\u5f8c\u5e36\u6709\u6b0a\u91cd\u7684\u76f8\u9130\u77e9\u9663\u3002" | |
| } | |
| ], | |
| "section": "\u9580\u63a7\u5716\u5e8f\u5217\u795e\u7d93\u7db2\u8def\u5c64 (GGSNN Layer)", | |
| "sec_num": "3.2" | |
| }, | |
| { | |
| "text": "EQUATION", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [ | |
| { | |
| "start": 0, | |
| "end": 8, | |
| "text": "EQUATION", | |
| "ref_id": "EQREF", | |
| "raw_str": "\u5728\u672c\u7814\u7a76\u7684\u9580\u63a7\u5716\u5e8f\u5217\u795e\u7d93\u7db2\u8def\u7d50\u69cb\u4e2d\uff0c\u7bc0\u9ede\u7684\u521d\u59cb\u72c0\u614b\u7531\u4ee5\u4e0b\u516c\u5f0f\u5f97\u5230\uff1a = \u2208 \u222a \u2208 (8) \u5176\u4e2d \u4ee3\u8868\u7684\u70ba\u591a\u91cd\u5d4c\u5165\u5c64\u6700\u5f8c\u8f38\u51fa\u7684\u5b57\u5e8f\u5217\u7279\u5fb5\u4e2d\uff0c\u6bcf\u500b\u5b57\u5206\u5225\u5c0d\u61c9\u5230\u7684\u7bc0\u9ede\uff0c \u5176\u503c\u7531\u591a\u91cd\u5d4c\u5165\u5c64\u6700\u5f8c\u8f38\u51fa\u7684\u5b57\u5e8f\u5217\u7279\u5fb5\u7684\u503c\u6c7a\u5b9a\uff0c \u70ba\u547d\u540d\u5be6\u9ad4\u7684\u8d77\u59cb\u5b57\u5c0d\u61c9\u5230\u7684\u7bc0 \u9ede\uff0c \u70ba\u547d\u540d\u5be6\u9ad4\u7684\u6700\u5f8c\u7684\u5b57\u5c0d\u61c9\u5230\u7684\u7bc0\u9ede\uff0c \u4ee5\u53ca \u7684\u503c\u70ba\u6bd4\u5c0d\u5230\u7684\u547d\u540d\u5be6\u9ad4\u7684\u96a8\u6a5f\u521d \u59cb\u72c0\u614b\u6c7a\u5b9a\u3002 \u7bc0\u9ede\u7684\u96b1\u85cf\u72c0\u614b\u85c9\u7531 GRU \u505a\u66f4\u65b0\uff0c\u6574\u500b\u905e\u8ff4\u95dc\u4fc2\u5f0f\u5982\u4e0b\uff1a \u57fa\u65bc\u5716\u795e\u7d93\u7db2\u8def\u4e4b\u4e2d\u6587\u5065\u5eb7\u7167\u8b77\u547d\u540d\u5be6\u9ad4\u8fa8\u8b58 29 = h , h , \u2026 \u2026 , h | | (9) = , \u2026 \u2026 , | | + b (10) = \u03c3 (11) = \u03c3 (12) = tanh \u2609 (13) = 1 \u2609 \u2609", | |
| "eq_num": "(14)" | |
| } | |
| ], | |
| "section": "\u9580\u63a7\u5716\u5e8f\u5217\u795e\u7d93\u7db2\u8def\u5c64 (GGSNN Layer)", | |
| "sec_num": "3.2" | |
| }, | |
| { | |
| "text": "\u5176\u4e2d \u8868\u793a\u7684\u70ba\u7bc0\u9ede \u5728\u6642\u9593\u70ba \u6642\u7684\u96b1\u85cf\u72c0\u614b\uff0c \u8868\u793a\u7684\u70ba\u7bc0\u9ede \u5c0d\u61c9\u7684\u76f8\u9130\u77e9 \u9663\u7684\u884c\u5411\u91cf\uff0c\u516c\u5f0f(11)-(14)\u70ba GRU \u55ae\u5143\uff0cz \u8207 r \u5206\u5225\u4ee3\u8868\u66f4\u65b0\u9580\u4ee5\u53ca\u91cd\u7f6e\u9580\uff0c\u900f\u904e GRU \u55ae\u5143\u53ef\u4ee5\u7d50\u5408\u4f86\u81ea\u76f8\u9130\u7bc0\u9ede\u7684\u4fe1\u606f\u4ee5\u53ca\u7bc0\u9ede\u7684\u7576\u524d\u96b1\u85cf\u72c0\u614b\uff0c\u8a08\u7b97\u5728\u6642\u9593 t \u6642\u65b0\u7684\u96b1\u85cf \u72c0\u614b\uff0c\u7d93\u904e\u6642\u9593\u6b65\u6578 (time step) T \u5f8c\uff0c\u53ef\u4ee5\u5f97\u5230\u7bc0\u9ede\u7684\u6700\u7d42\u72c0\u614b\u3002", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "\u9580\u63a7\u5716\u5e8f\u5217\u795e\u7d93\u7db2\u8def\u5c64 (GGSNN Layer)", | |
| "sec_num": "3.2" | |
| }, | |
| { | |
| "text": "\u5728\u9019\u5c64\u7d50\u69cb\u4e2d\uff0c\u672c\u7814\u7a76\u5c07\u9580\u63a7\u5716\u5e8f\u5217\u795e\u7d93\u7db2\u8def\u5c64\u6700\u5f8c\u7684\u96b1\u85cf\u72c0\u614b\u8f38\u51fa\uff0c\u7576\u4f5c BiLSTM \u8f38 \u5165\u5e8f\u5217\uff0c\u4f7f\u7528\u7684\u70ba\u6a19\u6e96\u7684 BiLSTM\u3002\u4ee5\u9580\u63a7\u5716\u5e8f\u5217\u795e\u7d93\u7db2\u8def\u5c64\u7684\u8f38\u51fa\u7576\u505a\u8f38\u5165\u5e8f\u5217\uff0c\u6700 \u7d42\u53ef\u4ee5\u7372\u5f97\u8207\u539f\u5e8f\u5217\u9577\u5ea6\u76f8\u540c\u7684\u96b1\u85cf\u5c64\u72c0\u614b\u5e8f\u5217\u3002 This work was partially supported by the Ministry of Science and Technology, Taiwan under the grant MOST 108-2218-E-008-017-MY3. We sincerely thank all the annotators for their efforts in the named entity tagging task. We also thank the anonymous reviewers for their insightful comments.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "\u96d9\u5411\u9577\u77ed\u671f\u8a18\u61b6\u795e\u7d93\u7db2\u8def\u5c64 (BiLSTM Layer)", | |
| "sec_num": "3.3" | |
| }, | |
| { | |
| "text": "\u570b\u5bb6\u7db2\u8def\u91ab\u85e5\uff1ahttps://www.kingnet.com.tw/knNew/index.html 2 \u5eb7\u5065\u96dc\u8a8c\uff1ahttps://www.commonhealth.com.tw/ 3 \u91ab\u806f\u7db2\uff1ahttps://med-net.com/", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "\u570b\u5bb6\u7db2\u8def\u91ab\u85e5\uff1ahttps://www.kingnet.com.tw/diagnose 5 \u570b\u5bb6\u6559\u80b2\u7814\u7a76\u9662\uff1ahttps://terms.naer.edu.tw/ 6 \u641c\u72d7\u7db2\uff1ahttps://pinyin.sogou.com/dict/", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "", | |
| "sec_num": null | |
| } | |
| ], | |
| "back_matter": [], | |
| "bib_entries": { | |
| "BIBREF0": { | |
| "ref_id": "b0", | |
| "title": "A coefficient of agreement for nominal scales", | |
| "authors": [ | |
| { | |
| "first": "J", | |
| "middle": [], | |
| "last": "Cohen", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1960, | |
| "venue": "Educational and Psychological Measurement", | |
| "volume": "20", | |
| "issue": "1", | |
| "pages": "37--46", | |
| "other_ids": { | |
| "DOI": [ | |
| "10.1177/001316446002000104" | |
| ] | |
| }, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Cohen, J. (1960). A coefficient of agreement for nominal scales. Educational and Psychological Measurement, 20(1), 37-46. doi: 10.1177/001316446002000104", | |
| "links": null | |
| }, | |
| "BIBREF1": { | |
| "ref_id": "b1", | |
| "title": "A neural multi-diagraph model for Chinese NER with gazetteers", | |
| "authors": [ | |
| { | |
| "first": "R", | |
| "middle": [], | |
| "last": "Ding", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "P", | |
| "middle": [], | |
| "last": "Xie", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "X", | |
| "middle": [], | |
| "last": "Zhang", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "W", | |
| "middle": [], | |
| "last": "Lu", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "L", | |
| "middle": [], | |
| "last": "Li", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "L", | |
| "middle": [], | |
| "last": "Si", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2019, | |
| "venue": "Proc. of the ACL", | |
| "volume": "19", | |
| "issue": "", | |
| "pages": "1462--1467", | |
| "other_ids": { | |
| "DOI": [ | |
| "10.18653/v1/P19-1141" | |
| ] | |
| }, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Ding, R., Xie, P., Zhang, X., Lu, W., Li, L., & Si, L. (2019). A neural multi-diagraph model for Chinese NER with gazetteers. In Proc. of the ACL'19, 1462-1467. doi: 10.18653/v1/P19-1141", | |
| "links": null | |
| }, | |
| "BIBREF2": { | |
| "ref_id": "b2", | |
| "title": "Character-based LSTM-CRF with radical-level features for Chinese named entity recognition", | |
| "authors": [ | |
| { | |
| "first": "C", | |
| "middle": [], | |
| "last": "Dong", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "J", | |
| "middle": [], | |
| "last": "Zhang", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "C", | |
| "middle": [], | |
| "last": "Zong", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "M", | |
| "middle": [], | |
| "last": "Hattori", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "H", | |
| "middle": [], | |
| "last": "Di", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2016, | |
| "venue": "Proc. of the ICCPOL", | |
| "volume": "16", | |
| "issue": "", | |
| "pages": "239--250", | |
| "other_ids": { | |
| "DOI": [ | |
| "10.1007/978-3-319-50496-4_20" | |
| ] | |
| }, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Dong, C., Zhang, J., Zong, C., Hattori, M., & Di, H. (2016). Character-based LSTM-CRF with radical-level features for Chinese named entity recognition. In Proc. of the ICCPOL'16, 239-250. doi: 10.1007/978-3-319-50496-4_20", | |
| "links": null | |
| }, | |
| "BIBREF3": { | |
| "ref_id": "b3", | |
| "title": "Measuring nominal scale agreement among many raters", | |
| "authors": [ | |
| { | |
| "first": "J", | |
| "middle": [ | |
| "L" | |
| ], | |
| "last": "Fleiss", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1971, | |
| "venue": "Psychological Bulletin", | |
| "volume": "76", | |
| "issue": "5", | |
| "pages": "378--382", | |
| "other_ids": { | |
| "DOI": [ | |
| ":10.1037/h0031619" | |
| ] | |
| }, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Fleiss, J. L. (1971). Measuring nominal scale agreement among many raters. Psychological Bulletin, 76(5), 378-382. doi: DOI: 10.1037/h0031619", | |
| "links": null | |
| }, | |
| "BIBREF4": { | |
| "ref_id": "b4", | |
| "title": "Conditional random fields: probabilistic models for segmenting and labeling sequence data", | |
| "authors": [ | |
| { | |
| "first": "J", | |
| "middle": [], | |
| "last": "Lafferty", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "A", | |
| "middle": [], | |
| "last": "Mccallum", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "F", | |
| "middle": [], | |
| "last": "Pereira", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2001, | |
| "venue": "Proc. of the ICML'01", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "282--289", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Lafferty, J., McCallum, A., & Pereira, F. (2001). Conditional random fields: probabilistic models for segmenting and labeling sequence data. In Proc. of the ICML'01, 282-289.", | |
| "links": null | |
| }, | |
| "BIBREF5": { | |
| "ref_id": "b5", | |
| "title": "Neural architectures for named entity recognition", | |
| "authors": [ | |
| { | |
| "first": "G", | |
| "middle": [], | |
| "last": "Lample", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "M", | |
| "middle": [], | |
| "last": "Ballesteros", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "S", | |
| "middle": [], | |
| "last": "Subramanian", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "K", | |
| "middle": [], | |
| "last": "Kawakami", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "C", | |
| "middle": [], | |
| "last": "Dyer", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2016, | |
| "venue": "Proc. of the NAACL-HLT'16", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "260--270", | |
| "other_ids": { | |
| "DOI": [ | |
| "10.18653/v1/N16-1030" | |
| ] | |
| }, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Lample, G., Ballesteros, M., Subramanian, S., Kawakami, K., & Dyer, C. (2016). Neural architectures for named entity recognition. In Proc. of the NAACL-HLT'16, 260-270. doi: 10.18653/v1/N16-1030", | |
| "links": null | |
| }, | |
| "BIBREF6": { | |
| "ref_id": "b6", | |
| "title": "The measurement of observer agreement for categorical data", | |
| "authors": [ | |
| { | |
| "first": "J", | |
| "middle": [ | |
| "R" | |
| ], | |
| "last": "Landis", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "G", | |
| "middle": [ | |
| "G" | |
| ], | |
| "last": "Koch", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1977, | |
| "venue": "Biometrics", | |
| "volume": "33", | |
| "issue": "1", | |
| "pages": "159--174", | |
| "other_ids": { | |
| "DOI": [ | |
| "10.2307/2529310" | |
| ] | |
| }, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Landis, J. R., & Koch, G. G. (1977). The measurement of observer agreement for categorical data. Biometrics, 33(1), 159-174. doi: 10.2307/2529310", | |
| "links": null | |
| }, | |
| "BIBREF7": { | |
| "ref_id": "b7", | |
| "title": "The third international Chinese language processing bakeoff: word segmentation and named entity recognition", | |
| "authors": [ | |
| { | |
| "first": "G", | |
| "middle": [ | |
| "A" | |
| ], | |
| "last": "Levow", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2006, | |
| "venue": "Proc. of the SIGHAN'06", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "108--117", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Levow, G. A. (2006). The third international Chinese language processing bakeoff: word segmentation and named entity recognition. In Proc. of the SIGHAN'06, 108-117.", | |
| "links": null | |
| }, | |
| "BIBREF8": { | |
| "ref_id": "b8", | |
| "title": "Gated graph sequence neural networks", | |
| "authors": [ | |
| { | |
| "first": "Y", | |
| "middle": [], | |
| "last": "Li", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "D", | |
| "middle": [], | |
| "last": "Tarlow", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "M", | |
| "middle": [], | |
| "last": "Brockschmidt", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "R", | |
| "middle": [], | |
| "last": "Zemel", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2016, | |
| "venue": "Proc. of the ICLR'16", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": { | |
| "arXiv": [ | |
| "arXiv:1511.05493.\u76e7\u6bc5\u8207\u674e\u9f8d\u8c6a" | |
| ] | |
| }, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Li, Y., Tarlow, D., Brockschmidt, M., & Zemel, R. (2016). Gated graph sequence neural networks. In Proc. of the ICLR'16. Retrieved from arXiv:1511.05493. \u76e7\u6bc5\u8207\u674e\u9f8d\u8c6a", | |
| "links": null | |
| }, | |
| "BIBREF9": { | |
| "ref_id": "b9", | |
| "title": "End-to-end sequence labeling via bi-directional LSTM-CNNs-CRF", | |
| "authors": [ | |
| { | |
| "first": "X", | |
| "middle": [], | |
| "last": "Ma", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "E", | |
| "middle": [], | |
| "last": "Hovy", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2016, | |
| "venue": "Proc. of the ACL", | |
| "volume": "16", | |
| "issue": "", | |
| "pages": "1064--1074", | |
| "other_ids": { | |
| "DOI": [ | |
| "10.18653/v1/P16-1101" | |
| ] | |
| }, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Ma, X., & Hovy, E. (2016). End-to-end sequence labeling via bi-directional LSTM-CNNs-CRF. In Proc. of the ACL'16, 1064-1074. doi: 10.18653/v1/P16-1101", | |
| "links": null | |
| }, | |
| "BIBREF10": { | |
| "ref_id": "b10", | |
| "title": "Named entity recognition for Chinese social media with jointly trained embeddings", | |
| "authors": [ | |
| { | |
| "first": "N", | |
| "middle": [], | |
| "last": "Peng", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "M", | |
| "middle": [], | |
| "last": "Dredze", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2015, | |
| "venue": "Proc. of EMNLP'15", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "548--554", | |
| "other_ids": { | |
| "DOI": [ | |
| "10.18653/v1/D15-1064" | |
| ] | |
| }, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Peng, N., & Dredze, M. (2015). Named entity recognition for Chinese social media with jointly trained embeddings. In Proc. of EMNLP'15, 548-554. doi: 10.18653/v1/D15-1064", | |
| "links": null | |
| }, | |
| "BIBREF11": { | |
| "ref_id": "b11", | |
| "title": "A tutorial on hidden markov models and selected applications in speech recognition", | |
| "authors": [ | |
| { | |
| "first": "L", | |
| "middle": [ | |
| "R" | |
| ], | |
| "last": "Rabiner", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1989, | |
| "venue": "Proc. of the IEEE", | |
| "volume": "77", | |
| "issue": "", | |
| "pages": "257--286", | |
| "other_ids": { | |
| "DOI": [ | |
| "10.1109/5.18626" | |
| ] | |
| }, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Rabiner, L. R. (1989). A tutorial on hidden markov models and selected applications in speech recognition. Proc. of the IEEE, 77(2), 257-286. doi: 10.1109/5.18626", | |
| "links": null | |
| }, | |
| "BIBREF12": { | |
| "ref_id": "b12", | |
| "title": "Enriching the knowledge sources used in a maximum entropy part-of-speech tagger", | |
| "authors": [ | |
| { | |
| "first": "K", | |
| "middle": [], | |
| "last": "Toutanova", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "C", | |
| "middle": [ | |
| "D" | |
| ], | |
| "last": "Manning", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2000, | |
| "venue": "Proc. of the EMNLP/VLC'00", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "63--70", | |
| "other_ids": { | |
| "DOI": [ | |
| "10.3115/1117794.1117802" | |
| ] | |
| }, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Toutanova, K., & Manning, C. D. (2000). Enriching the knowledge sources used in a maximum entropy part-of-speech tagger. In Proc. of the EMNLP/VLC'00, 63-70. doi: 10.3115/1117794.1117802", | |
| "links": null | |
| }, | |
| "BIBREF13": { | |
| "ref_id": "b13", | |
| "title": "Exploiting multiple embeddings for Chinese named entity recognition", | |
| "authors": [ | |
| { | |
| "first": "C", | |
| "middle": [], | |
| "last": "Xu", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "F", | |
| "middle": [], | |
| "last": "Wang", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "J", | |
| "middle": [], | |
| "last": "Han", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "C", | |
| "middle": [], | |
| "last": "Li", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2019, | |
| "venue": "Proc. of the CIKM", | |
| "volume": "19", | |
| "issue": "", | |
| "pages": "2269--2272", | |
| "other_ids": { | |
| "DOI": [ | |
| "10.1145/3357384.3358117" | |
| ] | |
| }, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Xu, C., Wang, F., Han, J., & Li, C. (2019). Exploiting multiple embeddings for Chinese named entity recognition. In Proc. of the CIKM'19, 2269-2272. doi: 10.1145/3357384.3358117", | |
| "links": null | |
| }, | |
| "BIBREF14": { | |
| "ref_id": "b14", | |
| "title": "Chinese NER using lattice LSTM", | |
| "authors": [ | |
| { | |
| "first": "Y", | |
| "middle": [], | |
| "last": "Zhang", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "J", | |
| "middle": [], | |
| "last": "Yang", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2018, | |
| "venue": "Proc. of the ACL", | |
| "volume": "18", | |
| "issue": "", | |
| "pages": "1554--564", | |
| "other_ids": { | |
| "DOI": [ | |
| "10.18653/v1/P18-1144" | |
| ] | |
| }, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Zhang, Y., & Yang, J. (2018). Chinese NER using lattice LSTM. In Proc. of the ACL'18, 1554-564. doi: 10.18653/v1/P18-1144", | |
| "links": null | |
| } | |
| }, | |
| "ref_entries": { | |
| "FIGREF0": { | |
| "uris": null, | |
| "num": null, | |
| "text": "\u5716 3. \u6709\u5411\u5716\u4ee5\u53ca\u5c0d\u61c9\u7684\u76f8\u9130\u77e9\u9663 [Figure 3. A directed graph and its corresponding adjacent matrix] \u76e7\u6bc5\u8207\u674e\u9f8d\u8c6a \u5716 4. \u591a\u7dad\u6709\u5411\u5716\u62c6\u89e3\u6210\u591a\u500b\u6709\u5411\u5716 [Figure 4. A directed multigraph is composed of multiple directed graphs]", | |
| "type_str": "figure" | |
| }, | |
| "TABREF1": { | |
| "text": "\u53e5\uff0c\u4eba\u5de5\u65b7\u8a5e\u4e26\u6a19\u8a18\u6642\u5f8c\uff0c\u900f\u904e\u8a08\u7b97 Cohen's Kappa \u503c\u4ee5\u53ca Fleiss' Kappa \u503c\u78ba\u4fdd\u6a19\u8a18\u7684\u54c1\u8cea\uff0c\u6700\u5f8c\u7e3d\u5171\u6709 68,460 \u500b\u547d\u540d\u5be6\u9ad4\uff0c\u6a6b\u8de8 10 \u500b\u985e \u5225\uff0c\u5206\u5225\u662f\u4eba\u9ad4\u3001\u75c7\u72c0\u3001\u91ab\u7642\u5668\u6750\u3001\u6aa2\u9a57\u3001\u5316\u5b78\u7269\u8cea\u3001\u75be\u75c5\u3001\u85e5\u54c1\u3001\u71df\u990a\u54c1\u3001\u6cbb\u7642\u4ee5\u53ca \u6642\u9593\u3002", | |
| "type_str": "table", | |
| "html": null, | |
| "num": null, | |
| "content": "<table><tr><td/><td>\u57fa\u65bc\u5716\u795e\u7d93\u7db2\u8def\u4e4b\u4e2d\u6587\u5065\u5eb7\u7167\u8b77\u547d\u540d\u5be6\u9ad4\u8fa8\u8b58</td><td>\u76e7\u6bc5\u8207\u674e\u9f8d\u8c6a 25</td></tr><tr><td colspan=\"3\">\u96c6\u7684\u8a13\u7df4\u8cc7\u6599\u7e3d\u5171\u5305\u542b\u4e86 46,364 \u500b\u53e5\u5b50\uff0c\u5176\u4e2d\u7684\u547d\u540d\u5be6\u9ad4\u7e3d\u6578\u70ba 118,643 \u500b\uff0c\u6e2c\u8a66\u8cc7\u6599 \u9664\u4e86\u5b57\u7684\u8cc7\u8a0a\u4ee5\u5916\uff0c\u672c\u7814\u7a76\u52a0\u5165\u4e86\u90e8\u9996\u4ee5\u53ca\u8a5e\u7684\u8cc7\u8a0a\u3002\u5728\u52a0\u5165\u5b57\u7279\u5fb5\u3001\u90e8\u9996\u7279\u5fb5\u6642\u900f\u904e</td></tr><tr><td colspan=\"3\">\u70ba 4,365 \u500b\u53e5\u5b50\uff0c\u5176\u4e2d\u6a19\u8a18\u7684\u547d\u540d\u5be6\u9ad4\u7e3d\u6578\u70ba 4,362 \u500b\u3002 \u96d9\u5411\u9577\u77ed\u671f\u8a18\u61b6\u985e\u795e\u7d93\u7db2\u8def\u4ee5\u53ca\u5377\u7a4d\u904b\u7b97\u505a\u4e86\u6709\u5225\u65bc Xu \u7b49\u4eba(2019)\u7684\u8655\u7406\uff0c\u4f7f\u7279\u5fb5\u8cc7\u8a0a</td></tr><tr><td colspan=\"3\">\u5728\u793e\u7fa4\u5a92\u9ad4\u65b9\u9762\uff0cWeibo \u547d\u540d\u5be6\u9ad4\u8a9e\u6599\u5eab(Peng & Dredze, 2015)\u8490\u96c6\u4e86\u5fae\u535a\u6b64\u793e\u7fa4\u5a92 \u66f4\u80fd\u5920\u5b8c\u6574\u5145\u5206\u3002\u5728\u672c\u7814\u7a76\u7684\u6a21\u578b\u540c\u6a23\u4f7f\u7528\u4e86\u6539\u826f\u5f0f\u7684 GGSNN \u5c07\u5b57\u5178\u8cc7\u8a0a\u52a0\u5165\uff0c\u800c\u8207</td></tr><tr><td colspan=\"3\">\u9ad4\u5f9e 2013 \u5e74 11 \u6708\u81f3 2014 \u5e74 12 \u6708\u7684\u8a0a\u606f\u4e26\u5c0d\u5176\u6a19\u8a18\uff0c\u96a8\u6a5f\u6311\u9078\u8a0a\u606f\u7684\u6578\u91cf\u7e3d\u5171\u70ba 1,890 Ding \u7b49\u4eba(2019)\u4e0d\u540c\u7684\u5730\u65b9\u5728\u65bc\u900f\u904e\u4e0d\u540c\u7684\u5b57\u5178\u7de8\u6392\u65b9\u5f0f\uff0c\u4f7f\u5176\u5728\u76f8\u540c\u7684\u786c\u9ad4\u8a2d\u5099\u4e0b\uff0c</td></tr><tr><td colspan=\"3\">\u5247\uff0c\u6a19\u8a18\u7684\u547d\u540d\u5be6\u9ad4\u985e\u5225\u7e3d\u5171\u6709 4 \u7a2e\uff0c\u5206\u5225\u70ba\u5730\u7406\u4f4d\u7f6e(Geo-political)\u3001\u5730\u540d(Location)\u3001 \u5b57\u5178\u7684\u4f86\u5b8c\u80fd\u5920\u66f4\u52a0\u7684\u9f90\u5927\u4e14\u8c50\u5bcc\u3002</td></tr><tr><td colspan=\"3\">\u7d44\u7e54\u540d(Organization)\u4ee5\u53ca\u4eba\u540d(Person)\uff0c\u5176\u4e2d\u6a19\u8a18\u7684\u547d\u540d\u5be6\u9ad4\u7e3d\u6578\u70ba 1,981 \u500b\u3002 3. \u6a21\u578b\u67b6\u69cb (Model Architecture)</td></tr><tr><td colspan=\"3\">Resume \u8cc7\u6599\u96c6(Zhang & Yang, 2018)\u7684\u4f86\u6e90\u70ba\u500b\u4eba\u5c65\u6b77\uff0c\u5c65\u6b77\u7684\u51fa\u8655\u70ba\u4e2d\u570b\u4e0a\u5e02\u516c</td></tr><tr><td colspan=\"3\">\u53f8\u4e3b\u7ba1\uff0c\u7e3d\u5171\u96a8\u6a5f\u6311\u9078\u4e86 1,027 \u4efd\uff0c\u6a19\u8a3b\u7684\u547d\u540d\u5be6\u9ad4\u7a2e\u985e\u7e3d\u5171\u6709 8 \u7a2e\uff0c\u5176\u4e2d\u5305\u542b\u570b\u5bb6</td></tr><tr><td colspan=\"3\">(Country)\u3001\u4eba\u540d(Person)\u4ee5\u53ca\u7d44\u7e54\u540d(Organization)\u7b49\u7b49\uff0c\u5176\u4e2d\u6a19\u8a18\u547d\u540d\u5be6\u9ad4\u7684\u7684\u7e3d\u6578\u70ba</td></tr><tr><td>16,565 \u500b\u3002</td><td>)\u3002</td></tr><tr><td colspan=\"3\">\u96a8\u8457\u79d1\u6280\u7684\u9032\u6b65\uff0c\u4eba\u985e\u7684\u58fd\u547d\u5f97\u4ee5\u5ef6\u9577\uff0c\u6709\u95dc\u5065\u5eb7\u7167\u8b77\u7684\u8b70\u984c\u9010\u6f38\u5730\u6d6e\u4e0a\u6aaf\u9762\uff0c\u8a31 \u4e2d\u570b\u77e5\u8b58\u5716\u8b5c\u8207\u8a9e\u7fa9\u8a08\u7b97\u5927\u6703(CCKS: China Conference on Knowledge Graph and</td></tr><tr><td colspan=\"3\">\u591a\u7684\u5831\u7ae0\u96dc\u8a8c\u90fd\u5728\u8ac7\u8ad6\u76f8\u95dc\u8b70\u984c\uff0c\u56e0\u6b64\u672c\u7814\u7a76\u6240\u95dc\u6ce8\u7684\u547d\u540d\u5be6\u9ad4\u9818\u57df\u9078\u5b9a\u70ba\u5065\u5eb7\u7167\u8b77\u3002 Semantic Computing)\u5728 2019 \u5e74\u8209\u8fa6\u7684\u8a55\u6e2c\u4efb\u52d9\u4e2d\uff0c\u547d\u540d\u5be6\u9ad4\u8fa8\u8b58\u7684\u8cc7\u6599\u4f86\u6e90\u70ba\u96fb\u5b50\u75c5\u6b77</td></tr><tr><td colspan=\"3\">\u6709\u9451\u65bc\u7576\u524d\u6b20\u7f3a\u5065\u5eb7\u7167\u8b77\u7684\u4e2d\u6587\u547d\u540d\u5be6\u9ad4\u8fa8\u8b58\u8a9e\u6599\u5eab\uff0c\u6211\u5011\u5f9e\u7db2\u8def\u4e0a\u8490\u96c6\u4e86\u76f8\u95dc\u7684\u6587\u7ae0 (Electronic Health Record, EHR)\uff0c\u8a13\u7df4\u96c6\u7684\u6587\u6a94\u6578\u70ba 1000 \u7b46\uff0c\u800c\u6e2c\u8a66\u96c6\u7684\u6587\u6a94\u6578\u70ba 379</td></tr><tr><td colspan=\"3\">\u7b46\uff0c\u6240\u6a19\u8a3b\u7684\u547d\u540d\u5be6\u9ad4\u5305\u542b 6 \u7a2e\uff0c\u5206\u5225\u70ba\u75be\u75c5\u548c\u8a3a\u65b7(Disease and Diagnosis)\u3001\u6aa2\u67e5 (Examination)\u4ee5\u53ca\u6aa2\u9a57(Inspection)\u7b49\u7b49\uff0c\u5176\u4e2d\u6a19\u8a18\u547d\u540d\u5be6\u9ad4\u7684\u7684\u7e3d\u6578\u70ba 16,565 \u500b\u3002 \u4e0a\u8ff0\u7684\u547d\u540d\u5be6\u9ad4\u8a9e\u6599\u5eab\uff0c\u4e26\u6c92\u6709\u95dc\u65bc\u5065\u5eb7\u7167\u8b77\u9818\u57df\u65b9\u9762\u7684\u8a9e\u6599\u5eab\uff0c\u4e14\u90fd\u70ba\u7c21\u9ad4\u4e2d\u6587\uff0c \u56e0\u6b64\u672c\u7814\u7a76\u5efa\u7f6e\u4e86\u4e00\u500b\u4e2d\u6587\u5065\u5eb7\u7167\u8b77\u9818\u57df\u7684\u547d\u540d\u5be6\u9ad4\u8a9e\u6599\u5eab\uff0c\u5171\u6709 10 \u985e\u547d\u540d\u5be6\u9ad4\uff0c\u5206\u5225 \u96dc\u8a8c\u4ee5\u53ca\u554f\u7b54\u7d00\u9304\uff0c\u96a8\u6a5f\u9078\u53d6 30,692 \u8fd1\u5e74\u4f86\u6df1\u5ea6\u5b78\u7fd2\u6280\u8853\u7684\u8208\u8d77\uff0c\u795e\u7d93\u7db2\u8def\u5728\u8a31\u591a\u4efb\u52d9\u7686\u6709\u8457\u4eae\u773c\u7684\u8868\u73fe\uff0c\u5728\u547d\u540d\u5be6\u9ad4 \u70ba\u4eba\u9ad4\u3001\u75c7\u72c0\u3001\u91ab\u7642\u5668\u6750\u3001\u6aa2\u9a57\u3001\u5316\u5b78\u7269\u8cea\u3001\u75be\u75c5\u3001\u85e5\u54c1\u3001\u71df\u990a\u54c1\u3001\u6cbb\u7642\u4ee5\u53ca\u6642\u9593\u3002</td></tr><tr><td colspan=\"3\">\u8fa8\u8b58\u4efb\u52d9\u4e2d\uff0cBiLSTM-CRF \u7db2\u8def\u67b6\u69cb\u70ba\u6700\u88ab\u5ee3\u6cdb\u4f7f\u7528\u7684\u4e3b\u6d41\u6a21\u578b(Lample, Ballesteros,</td></tr><tr><td colspan=\"3\">Subramanian, Kawakami & Dyer, 2016; Ma & Hovy, 2016)\u3002\u6211\u5011\u4ee5\u6b64\u67b6\u69cb\u70ba\u57fa\u790e\uff0c\u4e26\u4e14\u8003</td></tr><tr><td colspan=\"3\">\u91cf\u4e2d\u6587\u7684\u7279\u6027\uff0c\u65b7\u8a5e\u7684\u7cbe\u6e96\u5ea6\u6703\u56b4\u91cd\u5f71\u97ff\u7d50\u679c\uff0c\u56e0\u6b64\u4ee5\u5b57\u4f5c\u70ba\u8f38\u5165\u55ae\u4f4d\uff0c\u8a13\u7df4\u5b57\u5d4c\u5165\u3001 Dong \u7b49\u4eba(2016) \u8003\u91cf\u4e86\u4e2d\u6587\u5b57\u7684\u7279\u6027\uff0c\u5c07\u4e2d\u6587\u5b57\u62c6\u89e3\u6210\u4e00\u500b\u500b\u90e8\u4ef6\uff0c\u5176\u539f\u56e0\u70ba\u4e2d\u6587\u7684 \u90e8\u9996\u5d4c\u5165\u4ee5\u53ca\u8a5e\u5d4c\u5165\u8a9e\u610f\u5411\u91cf\uff0c\u900f\u904e\u9580\u63a7\u5716\u5e8f\u5217\u795e\u7d93\u7db2\u8def(Gated Graph Sequence Neural \u5b57\u662f\u7531\u8a31\u591a\u7684\u90e8\u4ef6\u7d44\u5408\u800c\u6210\uff0c\u800c\u6bcf\u500b\u90e8\u4ef6\u5177\u6709\u5176\u4e0d\u540c\u7684\u610f\u7fa9\uff0c\u900f\u904e\u9019\u4e9b\u90e8\u4ef6\u53ef\u4ee5\u589e\u52a0\u5b57</td></tr><tr><td colspan=\"3\">Networks, GGSNN)\u52a0\u5165\u5b57\u5178\u8cc7\u8a0a\uff0c\u5728\u5efa\u7f6e\u7684\u4e2d\u6587\u5065\u5eb7\u7167\u8b77\u547d\u540d\u5be6\u9ad4\u8fa8\u8b58\u8a9e\u6599\u5eab\u4e0a\uff0c\u9054\u5230 \u7279\u5fb5\u4ee5\u5916\u7684\u7279\u5fb5\uff0c\u5f9e\u8a72\u7814\u7a76\u53ef\u4ee5\u5f97\u77e5\u300c\u5b57\u300d\u4e26\u975e\u4e2d\u6587\u5b57\u5177\u6709\u610f\u7fa9\u7684\u6700\u5c0f\u55ae\u4f4d\u3002 F1 \u5206\u6578 75.69%\uff0c\u6bd4\u7576\u524d\u5177\u4ee3\u8868\u6027\u76f8\u95dc\u7814\u7a76\u6a21\u578b(BiLSTM-CRF, Lattice, Gazetteers \u4ee5\u53ca Xu \u7b49\u4eba(2019)\u52a0\u5165\u4e86\u9664\u4e86\u5b57\u7279\u5fb5\u4ee5\u5916\u7684\u90e8\u9996\u7279\u5fb5\u4ee5\u53ca\u8a5e\u7279\u5fb5\uff0c\u4e26\u4e14\u5c07\u5b57\u7279\u5fb5\u4ee5\u53ca\u90e8 ME-CNER)\u6709\u66f4\u597d\u7684\u6210\u6548\u3002 \u9996\u7279\u5fb5\u5229\u7528\u96d9\u5411\u9577\u77ed\u671f\u8a18\u61b6\u985e\u795e\u7d93\u7db2\u8def(BiLSTM)\u4ee5\u53ca\u5377\u7a4d\u904b\u7b97(Convolution)\u505a\u984d\u5916\u7684\u8655 \u672c\u7814\u7a76\u4e00\u5171\u5206\u70ba\u4e94\u500b\u7ae0\u7bc0\uff0c\u7b2c\u4e00\u7ae0\u7bc0\u70ba\u7dd2\u8ad6\uff0c\u4ecb\u7d39\u547d\u540d\u5be6\u9ad4\u8fa8\u8b58\u4efb\u52d9\u4ee5\u53ca\u7814\u7a76\u52d5\u6a5f \u7406\u3002\u5728\u6b64\u7814\u7a76\u4e2d\u4e4b\u6240\u4ee5\u52a0\u5165\u4e2d\u6587\u90e8\u9996\u7684\u539f\u56e0\u70ba\u4e2d\u6587\u90e8\u9996\u5177\u6709\u8a9e\u610f\u5206\u985e\uff0c\u540c\u6a23\u90e8\u9996\u7684\u5b57\uff0c \u8207\u76ee\u7684\u3002\u7b2c\u4e8c\u7ae0\u7bc0\u70ba\u63a2\u8a0e\u76f8\u95dc\u7814\u7a76\uff0c\u8abf\u67e5\u76ee\u524d\u7684\u4e2d\u6587\u547d\u540d\u5be6\u9ad4\u8fa8\u8b58\u8a9e\u6599\u5eab\uff0c\u4e26\u4e14\u4ecb\u7d39\u4e2d \u53ef\u80fd\u5c6c\u65bc\u540c\u6a23\u985e\u5225\uff0c\u56e0\u6b64\u900f\u904e\u90e8\u9996\u53ef\u4ee5\u5c0d\u5b57\u505a\u66f4\u9032\u4e00\u6b65\u7684\u5206\u6790\u3002 \u6587\u547d\u540d\u5be6\u9ad4\u8fa8\u8b58\u6a21\u578b\u3002\u7b2c\u4e09\u7ae0\u7bc0\u70ba\u6a21\u578b\u67b6\u69cb\uff0c\u8a73\u7d30\u4ecb\u7d39\u63d0\u51fa\u7684\u5716\u795e\u7d93\u7db2\u8def\u6a21\u578b\uff0c\u4e26\u5c0d\u6a21 Zhang \u548c Yang (2018)\u63d0\u51fa\u4e86\u4e00\u500b\u65b0\u7684\u6a21\u67b6\u69cb Lattice LSTM\uff0c\u6b64\u6a21\u578b\u4e3b\u8981\u7684\u7279\u9ede\u70ba\u6703 \u578b\u7684\u5404\u5c64\u505a\u8a73\u76e1\u7684\u8aaa\u660e\u3002\u7b2c\u56db\u7ae0\u7bc0\u70ba\u5be6\u9a57\u8a55\u4f30\u8207\u5206\u6790\uff0c\u4f9d\u5e8f\u8aaa\u660e\u8a9e\u6599\u5eab\u7684\u5efa\u7f6e\u3001\u5d4c\u5165\u5411 \u5c07\u53e5\u5b50\u4e2d\u8a5e\u5f59\u900f\u904e\u5927\u578b\u81ea\u52d5\u53d6\u5f97\u7684\u5b57\u5178\uff0c\u5c07\u6240\u6709\u53ef\u80fd\u7684\u6f5b\u5728\u8a5e\u5f59\u627e\u51fa\uff0c\u5229\u7528\u6b64\u7a2e\u65b9\u5f0f\u53ef \u91cf\u3001\u5be6\u9a57\u8a2d\u5b9a\u8207\u6548\u80fd\u8a55\u4f30\u6307\u6a19\uff0c\u63a5\u8457\u8a0e\u8ad6\u5be6\u9a57\u7d50\u679c\u548c\u932f\u8aa4\u5206\u6790\u3002\u7b2c\u4e94\u7ae0\u70ba\u7d50\u8ad6\u548c\u672a\u4f86\u7814 \u4ee5\u5c07\u8003\u91cf\u5230\u53ef\u80fd\u6f5b\u5728\u7684\u8a5e\u908a\u754c\uff0c\u6b64\u7814\u7a76\u7d50\u679c\u5728\u547d\u540d\u5be6\u9ad4\u8fa8\u8b58\u7684\u4efb\u52d9\u4e2d\u53d6\u5f97\u4e86\u91cd\u5927\u7684\u6210 \u7a76\u3002 \u679c\u3002</td></tr><tr><td colspan=\"3\">Ding \u7b49\u4eba(2019)\u4f7f\u7528\u5230\u4e86\u5716\u795e\u7d93\u7db2\u8def\u4e2d\u7684\u9580\u63a7\u5716\u5e8f\u5217\u795e\u7d93\u7db2\u8def\uff0c\u4e26\u505a\u6539\u826f\u4f7f\u5176\u80fd\u5920 \u5c07\u591a\u500b\u5b57\u5178\u7684\u8cc7\u8a0a\u52a0\u5165\u6a21\u578b\uff0c\u7531\u65bc\u6587\u5b57\u8a0a\u606f\u5e38\u5e38\u6703\u6709\u8457\u985e\u4f3c\u5716\u7d50\u69cb\u7684\u8a0a\u606f\uff0c\u56e0\u6b64\u900f\u904e\u5716 \u795e\u7d93\u7db2\u8def\u80fd\u5920\u66f4\u5145\u5206\u7684\u8868\u9054\u8cc7\u8a0a\u3002 2. \u76f8\u95dc\u7814\u7a76 MSRA \u547d\u540d\u5be6\u9ad4\u8a9e\u6599\u5eab(Levow, 2006)\u7e3d\u5171\u5305\u542b 30 \u7a2e\u985e\u5225\uff0c\u8a9e\u6599\u4f86\u6e90\u70ba\u65b0\u805e\u6587\u7ae0\uff0c\u5176\u4e2d\u8f03 \u57fa\u65bc\u4e0a\u8ff0\u7814\u7a76\uff0c\u672c\u7814\u7a76\u63d0\u51fa\u4e86\u9580\u63a7\u5716\u5e8f\u5217\u795e\u7d93\u7db2\u8def(Gated Graph Sequence Neural</td></tr><tr><td colspan=\"3\">\u88ab\u5ee3\u6cdb\u4f7f\u7528\u7684\u985e\u5225\u50cf\u662f\u4eba\u540d(Person)\u3001\u5730\u540d(Location \u4ee5\u53ca\u7d44\u7e54\u540d(Organization)\uff0c\u6b64\u8cc7\u6599 Networks, GGSNN)\u6a21\u578b\u67b6\u69cb\uff0c\u4ee5 BiLSTM-CRF \u505a\u70ba\u57fa\u790e\uff0c\u4ee5\u5b57\u70ba\u55ae\u4f4d\u7576\u4f5c\u6a21\u578b\u7684\u8f38\u5165\u3002</td></tr></table>" | |
| }, | |
| "TABREF2": { | |
| "text": "\u3002\u5176\u4e2d \u70ba\u5b57\u5e8f\u5217\u7bc0\u9ede\u7684\u96c6\u5408\uff0c\u800c\u7576\u5b57\u5178\u6bd4\u5c0d \u5230\u8a5e\u5f59\u6642\uff0c\u6703\u7522\u751f\u9664\u4e86\u5b57\u5e8f\u5217\u7bc0\u7684\u984d\u5916\u5169\u500b\u7bc0\u9ede\uff0c\u5206\u5225\u70ba , \u3001 , \uff0c\u5176\u4e2d , \u6307\u793a\u51fa\u8a5e \u5f59\u7684\u8d77\u59cb\u4f4d\u7f6e\uff0c , \u6307\u793a\u51fa\u8a5e\u5f59\u7684\u7d50\u675f\u4f4d\u7f6e\uff0c \u4ee5\u53ca \u5206\u5225\u4ee3\u8868\u7684\u70ba , \u4ee5\u53ca , \u7684\u96c6\u5408\u3002 \u908a\u7684\u96c6\u5408", | |
| "type_str": "table", | |
| "html": null, | |
| "num": null, | |
| "content": "<table><tr><td colspan=\"3\">\u57fa\u65bc\u5716\u795e\u7d93\u7db2\u8def\u4e4b\u4e2d\u6587\u5065\u5eb7\u7167\u8b77\u547d\u540d\u5be6\u9ad4\u8fa8\u8b58</td><td/><td>27</td></tr><tr><td colspan=\"5\">\u6709\u5411\u5716\uff0c\u5728\u6b64\u53e5\u5b50\u4e2d\uff0c\u53ef\u4ee5\u6bd4\u5c0d\u5230\u7684\u8a5e\u5f59\u6709\u300c\u601d\u89ba\u5931\u8abf\u75c7\u300d\u3001\u300c\u5931\u8abf\u300d\u3001\u300c\u5927\u8166\u300d\u4ee5\u53ca</td></tr><tr><td colspan=\"5\">\u300c\u591a\u5df4\u80fa\u300d\uff0c\u5176\u4e2d\u300c\u601d\u89ba\u5931\u8abf\u75c7\u300d\u5305\u542b\u5728\u8a5e\u5f59\u5b57\u6578\u70ba 5 \u500b\u5b57\u4ee5\u4e0a (else) \u7684\u5b57\u5178\u4e2d\uff0c\u56e0\u6b64</td></tr><tr><td colspan=\"3\">\u300c\u601d\u89ba\u5931\u8abf\u75c7\u300d\u7684\u958b\u982d\u300c\u601d\u300d\uff0c\u5c0d\u61c9\u5230\u7684\u7bc0\u9ede \uff0c\u9023\u7d50\u4e86\u984d\u5916\u7684\u7bc0\u9ede</td><td colspan=\"2\">, \uff0c\u300c\u601d\u89ba\u5931</td></tr><tr><td colspan=\"2\">\u8abf\u75c7\u300d\u7684\u7d50\u5c3e\u300c\u75c7\u300d\uff0c\u5c0d\u61c9\u5230\u7684\u7bc0\u9ede \uff0c\u9023\u7d50\u4e86\u984d\u5916\u7684\u7bc0\u9ede</td><td>, \uff0c</td><td colspan=\"2\">, \u7684\u4e0b\u6a19</td><td>\u4ee5</td></tr><tr><td colspan=\"2\">\u53ca\u4e0b\u6a19 s \u4ee3\u8868\u7684\u70ba\u6bd4\u5c0d\u5230\u7684\u5b57\u5178\u4ee5\u53ca\u6bd4\u5c0d\u5230\u7684\u8a5e\u7684\u958b\u982d\u4f4d\u7f6e\uff0c</td><td colspan=\"2\">, \u7684\u4e0b\u6a19</td><td>\u4ee5\u53ca\u4e0b</td></tr><tr><td colspan=\"4\">\u6a19 e \u4ee3\u8868\u7684\u70ba\u6bd4\u5c0d\u5230\u7684\u5b57\u5178\u4ee5\u53ca\u6bd4\u5c0d\u5230\u7684\u8a5e\u7684\u7d50\u5c3e\u4f4d\u7f6e\uff0c\u5176\u9918\u4f9d\u6b64\u985e\u63a8\u3002</td></tr><tr><td/><td>\u5716 2. \u591a\u7dad\u6709\u5411\u5716</td><td/><td/></tr><tr><td/><td>[Figure 2. A directed multigraphs]</td><td/><td/></tr><tr><td colspan=\"5\">\u6709\u5411\u5716\u7684\u7d50\u69cb\u8a0a\u606f\uff0c\u53ef\u4ee5\u900f\u904e\u76f8\u9130\u77e9\u9663 (adjacency matrix) \u8868\u9054\uff0c\u5047\u8a2d\u6709\u5411\u5716\u7684\u7d50</td></tr><tr><td colspan=\"3\">\u69cb\u70ba\u5716 3 \u7684\u5de6\u534a\u90e8\uff0c\u800c\u5176\u5c0d\u61c9\u7684\u76f8\u9130\u77e9\u9663\u5982\u5716 3 \u7684\u53f3\u534a\u90e8\uff0c\u5176\u4e2d \u8207</td><td colspan=\"2\">\u4e92\u70ba\u8f49\u7f6e\u77e9\u9663\uff0c</td></tr><tr><td>\u800c\u76f8\u9130\u77e9\u9663\u7531 \u4ee5\u53ca</td><td>\u6240\u69cb\u6210\u3002</td><td/><td/></tr><tr><td colspan=\"5\">\u4ee5\u7bc4\u4f8b\u53e5\u5b50\u300c\u601d\u89ba\u5931\u8abf\u75c7\u8207\u5927\u8166\u7684\u591a\u5df4\u80fa\u6709\u95dc\u300d\u70ba\u4f8b\uff0c\u8a72\u53e5\u5b50\u7684\u591a\u7dad\u6709\u5411\u5716\u7684\u62c6\u89e3</td></tr><tr><td colspan=\"5\">\u6210\u591a\u500b\u6709\u5411\u5716\u7684\u7bc4\u4f8b\u5982\u5716 4\uff0c\u7531\u65bc\u8a5e\u5f59\u5b57\u6578\u70ba 1 \u500b\u5b57\u7684\u5b57\u5178\u4ee5\u53ca\u8a5e\u5f59\u5b57\u6578\u70ba 4 \u500b\u5b57\u7684\u5b57</td></tr><tr><td colspan=\"2\">\u5178\u4e26\u6c92\u6709\u6bd4\u5c0d\u5230\u8a5e\u5f59\uff0c\u56e0\u6b64\u5c0d\u61c9\u7684\u76f8\u9130\u77e9\u9663\u70ba\u96f6\u77e9\u9663\u3002</td><td/><td/></tr><tr><td/><td/><td colspan=\"3\">\uff0c\u5176\u4e2d \u4ee3\u8868\u7bc0\u9ede\u7684</td></tr><tr><td colspan=\"5\">\u96c6\u5408\uff0c \u4ee3\u8868\u908a\u7684\u96c6\u5408\uff0c \u4ee3\u8868\u908a\u4e0a\u6a19\u7c64\u7684\u96c6\u5408\u3002\u5047\u8a2d\u8f38\u5165\u7684\u53e5\u5b50\u70ba\u5b57\u6578\u70ba n \u500b\uff0c\u5b57\u5178\u7684</td></tr><tr><td colspan=\"3\">\u4f7f\u7528\u6578\u91cf\u70ba m\uff0c\u7bc0\u9ede\u7684\u96c6\u5408 \u222a \u222a \u222a \uff0c\u5176\u4e2d \u70ba\u5b57\u5e8f\u5217\u7bc0\u9ede\u9023\u6210\u7684\u908a\u7684\u96c6\u5408\uff0c</td><td colspan=\"2\">\u70ba\u6240\u6709\u5b57\u5178</td></tr><tr><td colspan=\"2\">\u9023\u6210\u7684\u908a\u7684\u96c6\u5408\u3002\u6bcf\u500b\u908a\u90fd\u5e36\u6709\u6a19\u7c64\uff0c\u908a\u4e0a\u6a19\u7c64\u7684\u96c6\u5408\u70ba</td><td>\u222a</td><td colspan=\"2\">\uff0c \u70ba\u5b57\u5e8f\u5217</td></tr><tr><td colspan=\"5\">\u7bc0\u9ede\u9023\u6210\u7684\u908a\u4e0a\u7684\u6a19\u7c64\uff0c \u70ba\u5b57\u5178\u9023\u6210\u7684\u908a\u4e0a\u7684\u6a19\u7c64\uff0c\u4e0d\u540c\u7684\u5b57\u5178\u5e36\u6709\u4e0d\u540c\u7684\u6a19\u7c64\u3002</td></tr><tr><td colspan=\"5\">\u4ee5\u300c\u601d\u89ba\u5931\u8abf\u75c7\u8207\u5927\u8166\u7684\u591a\u5df4\u80fa\u6709\u95dc\u300d\u7576\u4f5c\u8f38\u5165\u53e5\u5b50\u70ba\u4f8b\uff0c\u53ef\u4ee5\u5f97\u5230\u5982\u5716 2 \u7684\u591a\u7dad</td></tr></table>" | |
| }, | |
| "TABREF3": { | |
| "text": "\u672c\u7814\u7a76\u900f\u904e\u722c\u87f2\u5c07\u7db2\u8def\u4e0a\u7684\u5065\u5eb7\u7167\u8b77\u6587\u7ae0\u53ca\u554f\u7b54\u7d00\u9304\u722c\u53d6\u4e0b\u4f86\uff0c\u6709\u4e09\u7a2e\u4f86\u6e90\u5206\u5225\u70ba\u570b\u5bb6 \u7db2\u8def\u91ab\u85e5 1 \u3001\u5eb7\u5065\u96dc\u8a8c 2 \u548c\u91ab\u806f\u7db2 3 \u3002\u5176\u4e2d\u570b\u5bb6\u7db2\u8def\u91ab\u85e5\u4ee5\u53ca\u5eb7\u5065\u96dc\u8a8c\u70ba\u91ab\u751f\u6216\u662f\u76f8\u95dc\u7684\u5c08 \u696d\u4eba\u54e1\u6240\u64b0\u5beb\u7684\u6587\u7ae0\uff0c\u800c\u91ab\u806f\u7db2\u5247\u662f\u4e00\u822c\u6c11\u773e\u4e0a\u7db2\u63d0\u554f\uff0c\u91ab\u751f\u56de\u7b54\u7684\u554f\u7b54\u7d00\u9304\uff0c\u6587\u7ae0\u5167 \u5bb9\u900f\u904e\u7be9\u9078\u4e3b\u984c\u9078\u64c7\u5065\u5eb7\u7167\u8b77\u76f8\u95dc\u3002\u672c\u7814\u7a76\u5206\u5225\u5728\u570b\u5bb6\u7db2\u8def\u91ab\u85e5\u4ee5\u53ca\u5eb7\u5065\u96dc\u8a8c\u4e00\u5171\u722c\u53d6 \u4e86 425 \u7bc7\u6587\u7ae0\u4ee5\u53ca 799 \u7bc7\u6587\u7ae0\uff0c\u800c\u91ab\u7642\u7db2\u4e00\u5171\u6709 1,818 \u5247\u554f\u7b54\u3002 \u900f\u904e\u8a08\u7b97 Cohen's Kappa(Cohen, 1960) \u503c\u4ee5\u53ca Fleiss' Kappa(Fleiss, 1971) \u503c\u53ef\u4ee5\u78ba \u4fdd\u6a19\u8a18\u54c1\u8cea\uff0c\u5176\u4e3b\u8981\u7684\u529f\u80fd\u70ba\u8a55\u4f30\u554f\u984c\u7684\u4e00\u81f4\u6027\uff0c\u5176\u4e2d Cohen's Kappa \u503c\u9069\u7528\u65bc\u6aa2\u5b9a\u5169 \u500b\u4eba\u610f\u898b\u7684\u4e00\u81f4\u6027\uff0c\u800c Fleiss' Kappa \u503c\u5247\u7528\u4f86\u6aa2\u5b9a\u4e09\u4eba\u4ee5\u4e0a\u7684\u60c5\u6cc1\u3002\u6839\u64da Landis \u4ee5\u53ca 30 \u76e7\u6bc5\u8207\u674e\u9f8d\u8c6a Koch \u6240\u63d0\u51fa\u7684\u89c0\u9ede (Landis & Koch, 1977)\uff0c\u7576 Kappa \u503c\u5c0f\u65bc 0 \u6642\u70ba Poor agreement\uff0c\u4ecb \u65bc 0 \u5230 0.20 \u70ba Slight agreement\uff0c\u4ecb\u65bc 0.21 \u5230 0.40 \u70ba Fair agreement\uff0c\u4ecb\u65bc 0.41 -0.60 \u70ba Moderate agreement\uff0c\u4ecb\u65bc 0.61 -0.80 \u70ba Substantial agreement\uff0c\u4ecb\u65bc 0.81 -1.00 F1\uff0c\u56e0\u6b64\u900f\u904e GGSNN \u5c07\u5b57\u5178\u7684\u8cc7\u8a0a\u7d0d\u5165\u8003\u616e\uff0c\u53ef\u4ee5\u6709\u6548\u7684\u63d0\u5347\u6a21\u578b\u7684\u8868\u73fe\u3002 \u672c\u7814\u7a76\u63d0\u51fa\u7684 GGSNN \u6a21\u578b\uff0c\u540c\u6642\u52a0\u5165\u4e86\u90e8\u9996\u5d4c\u5165\u3001\u8a5e\u5d4c\u5165\u4ee5\u53ca\u5716\u5e8f\u5217\u795e\u7d93\u7db2\u8def\uff0c\u5176\u8868 \u73fe\u8207 ME-CNER \u4ee5\u53ca Gazetteers \u6bd4\u8f03\uff0c\u5206\u5225\u4e0a\u5347\u4e86 1.54 \u4ee5\u53ca 1.43 \u7684 F1\u3002 \u7531 GGSNN \u5206\u5225\u53bb\u9664\u6389\u90e8\u9996\u5d4c\u5165\u3001\u8a5e\u5d4c\u5165\u4ee5\u53ca\u540c\u6642\u53bb\u9664\u5169\u8005\u7684\u5be6\u9a57\u6bd4\u8f03\u4e2d\uff0c\u53ef\u4ee5\u66f4 \u52a0\u5730\u78ba\u8a8d\u90e8\u9996\u5d4c\u5165\u4ee5\u53ca\u8a5e\u5d4c\u5165\u5c0d\u65bc\u6a21\u578b\u7684\u8868\u73fe\u5f71\u97ff\uff0c\u53bb\u9664\u90e8\u9996\u5d4c\u5165\u6a21\u578b\u7684 F1-score \u4e0b\u964d \u4e86 0.61\uff0c\u53bb\u9664\u8a5e\u5d4c\u5165\u6a21\u578b\u7684 F1-score \u4e0b\u964d\u4e86 1.41\uff0c\u540c\u6642\u53bb\u9664\u5169\u8005\u6a21\u578b\u7684 F1-score \u4e0b\u964d\u4e86 1.69\uff0c\u56e0\u6b64\u6211\u5011\u53ef\u4ee5\u5f97\u77e5\u8a5e\u5d4c\u5165\u5c0d\u65bc\u63d0\u5347\u6a21\u578b\u7684\u8868\u73fe\u7684\u8ca2\u737b\u8f03\u5927\uff0c\u800c\u4e0d\u8ad6\u662f\u8a5e\u5d4c\u5165\u6216\u662f \u90e8\u9996\u5d4c\u5165\uff0c\u7686\u5c0d\u6a21\u578b\u7684\u8868\u73fe\u6709\u5e6b\u52a9\u3002 \u672c\u7814\u7a76\u63d0\u51fa\u7684 GGSNN \u6a21\u578b\u6709\u6700\u4f73\u7684 F1 \u5206\u6578\uff0c\u800c Lattice \u6b21\u4e4b\uff0c\u5169\u500b\u6a21\u578b\u7684\u5dee\u7570\u70ba \u7565\u5c0f\u53ea\u6709 0.47\uff0c\u7136\u800c\u5728\u8a13\u7df4\u7684\u6642\u9593\u65b9\u9762\uff0c\u76f8\u540c\u7684\u786c\u9ad4\u8a2d\u5099\u4e0b\uff0c\u672c\u7814\u7a76\u7684\u6a21\u578b\u7d04\u70ba 1 \u5929\uff0c \u800c Lattice \u7d04\u8017\u6642 6.25 \u5929\uff0c\u4e3b\u8981\u7684\u539f\u56e0\u70ba Lattice \u6a21\u578b\u7684 batch size \u56e0\u70ba\u6a21\u578b\u7684\u7279\u6027\u53ea\u80fd\u5920 \u8a2d\u5b9a\u70ba 1\uff0c\u7576\u8cc7\u6599\u91cf\u8d8a\u5927\u6642\uff0c\u9700\u8981\u66f4\u591a\u6642\u9593\uff0c\u7121\u6cd5\u85c9\u7531\u8abf\u6574 batch size \u52a0\u901f\u904b\u7b97\u3002 \u8868", | |
| "type_str": "table", | |
| "html": null, | |
| "num": null, | |
| "content": "<table><tr><td>32</td><td>\u57fa\u65bc\u5716\u795e\u7d93\u7db2\u8def\u4e4b\u4e2d\u6587\u5065\u5eb7\u7167\u8b77\u547d\u540d\u5be6\u9ad4\u8fa8\u8b58 \u57fa\u65bc\u5716\u795e\u7d93\u7db2\u8def\u4e4b\u4e2d\u6587\u5065\u5eb7\u7167\u8b77\u547d\u540d\u5be6\u9ad4\u8fa8\u8b58 \u57fa\u65bc\u5716\u795e\u7d93\u7db2\u8def\u4e4b\u4e2d\u6587\u5065\u5eb7\u7167\u8b77\u547d\u540d\u5be6\u9ad4\u8fa8\u8b58</td><td>31 \u76e7\u6bc5\u8207\u674e\u9f8d\u8c6a 33 \u76e7\u6bc5\u8207\u674e\u9f8d\u8c6a 35</td></tr><tr><td colspan=\"3\">\u6574\u500b\u8a9e\u6599\u5eab\u6700\u5f8c\u5305\u542b 30,692 \u53e5\uff0c\u7e3d\u5b57\u6578\u7d04 150 \u842c\u5b57\uff0c\u63a5\u8fd1 92 \u842c\u500b\u8a5e\uff0c68,640 \u500b\u547d\u540d\u5be6 \u5360\u300c\u61c9\u8a72\u88ab\u8fa8\u8b58\u7684\u9805\u76ee\u300d\u7684\u6bd4\u4f8b\u4ee5\u53ca F1-score \u6b64\u70ba Precision \u4ee5\u53ca Recall \u7684\u8abf\u548c\u5e73\u5747\u6578\uff0c \u7528\u5230\u7684\u5b57\u5d4c\u5165\u4ee5\u53ca\u8a5e\u5d4c\u5165\u7531\u958b\u6e90\u7a0b\u5f0f\u78bc\u6240\u63d0\u4f9b\u3002 4.6 \u932f\u8aa4\u5206\u6790 (Error Analysis) \u4e8c\u3001\u64da\u6211\u5011\u6240\u77e5\uff0c\u9019\u662f\u7b2c\u4e00\u500b\u5065\u5eb7\u7167\u8b77\u9818\u57df\u7684\u4e2d\u6587\u547d\u540d\u5be6\u9ad4\u8a9e\u6599\u5eab\uff0c\u5305\u542b 30,692 \u500b\u53e5\u5b50\uff0c</td></tr><tr><td colspan=\"3\">\u9ad4\u3002\u8a13\u7df4\u8cc7\u6599\u662f\u4e09\u500b\u6a19\u8a18\u4eba\u54e1\u5404\u81ea\u6a19\u8a18\u7684\u90e8\u5206\uff0c\u5171\u6709 28,161 \u53e5\uff0c\u6bcf\u500b\u53e5\u5b50\u5e73\u5747 49.44 \u500b \u8a08\u7b97\u516c\u5f0f\u5982\u65b9\u7a0b\u5f0f (15)-(17)\u3002 (5)\u3001GGSNN\uff1a \u672c\u7814\u7a76\u5c07\u547d\u540d\u5be6\u9ad4\u7684\u932f\u8aa4\u5206\u6210\u4ee5\u4e0b 5 \u7a2e\u985e\u578b\uff0c\u932f\u8aa4\u7bc4\u4f8b\u5982\u8868 4\u3002 \u7d04\u83ab 150 \u842c\u5b57 (92 \u842c\u8a5e)\uff0c\u5171\u6709 68,460 \u500b\u547d\u540d\u5be6\u9ad4\uff0c\u6a6b\u8de8 10 \u500b\u985e\u5225\uff0c\u5305\u542b\uff1a\u4eba\u9ad4\u3001\u75c7</td></tr><tr><td colspan=\"3\">\u547d\u540d\u5be6\u9ad4\u8fa8\u8b58\u5c6c\u65bc\u5e8f\u5217\u6a19\u8a18\u7684\u591a\u5206\u985e\u554f\u984c\uff0c\u50b3\u7d71\u4e0a\u5728\u9047\u5230\u591a\u5206\u985e\u554f\u984c\u6642\uff0c\u6703\u63a1\u7528 softmax function \u4f5c\u70ba\u8f38\u51fa\u51fd\u6578\uff0c\u4f46\u5728\u5be6\u969b\u60c5\u6cc1\u6642\uff0c\u5e8f\u5217\u6a19\u8a3b\u4efb\u52d9\u4e2d\u7684\u7576\u524d\u6642\u523b\u7684\u72c0\u614b\uff0c\u5747\u8207\u7576 \u524d\u6642\u523b\u7684\u524d\u5f8c\u72c0\u614b\u6709\u6240\u95dc\u9023\uff0c\u56e0\u6b64\u689d\u4ef6\u96a8\u6a5f\u5834\u57df (Condition Random Fields, CRF) \u53d6\u4ee3\u4e86 softmax function\uff0c\u6210\u70ba\u4e86\u7576\u524d\u4e3b\u6d41\u7684\u67b6\u69cb\uff0c\u672c\u7814\u7a76\u6240\u63a1\u7528\u7684\u70ba\u6a19\u6e96\u7684 CRF \u6a21\u578b\u3002 \u70ba Almost perfect agreement\u3002 \u672c\u7814\u7a76\u6240\u95dc\u6ce8\u7684\u5065\u5eb7\u7167\u8b77\u547d\u540d\u5be6\u9ad4\uff0c\u7e3d\u5171\u5305\u542b 10 \u985e\uff0c\u5176\u5b9a\u7fa9\u4ee5\u53ca\u4f8b\u5b50\u5982\u8868 1\u3002\u6574\u500b \u6a19\u8a18\u8cc7\u6599\u7684\u6d41\u7a0b\uff0c\u6211\u5011\u5c07\u5176\u5206\u6210\u5169\u500b\u968e\u6bb5\uff0c\u53c3\u8207\u6a19\u8a18\u7684\u4eba\u54e1\u4e00\u5171\u6709\u4e09\u4f4d\u5e2b\u5927\u4e2d\u6587\u7cfb\u7684\u5927 \u5b78\u751f\uff0c\u5c0d\u65bc\u6bcf\u500b\u4e2d\u6587\u53e5\u5b50\u505a\u4eba\u5de5\u65b7\u8a5e\u53ca\u547d\u540d\u5be6\u9ad4\u6a19\u8a18\uff0c\u7b2c\u4e00\u500b\u968e\u6bb5\u5148\u53d6\u570b\u5bb6\u7db2\u8def\u91ab\u85e5 25 \u7bc7\u6587\u7ae0\u3001\u5eb7\u5065\u96dc 25 \u7bc7\u6587\u7ae0\u4ee5\u53ca\u91ab\u806f\u7db2 100 \u5247\u554f\u7b54\uff0c\u5148\u505a\u7b2c\u4e00\u6b21\u6a19\u8a18\uff0c\u8a08\u7b97\u4e09\u4f4d\u6a19\u8a18\u4eba\u54e1 \u7684\u4e00\u81f4\u6027\uff0c\u5f97\u5230 Fleiss' Kappa \u503c\u70ba 0.80\u3002\u5c0d\u968e\u6bb5\u4e00\u7684\u6a19\u8a18\u7d50\u679c\u505a\u8a0e\u8ad6\uff0c\u4fee\u6b63\u6a19\u8a18\u6e96\u5247\u5f97 \u5230\u4e00\u81f4\u7684\u6a19\u6e96\u5f8c\uff0c\u518d\u5c0d\u53e6\u5916\u7684 25 \u7bc7\u570b\u5bb6\u7db2\u8def\u91ab\u85e5\u7684\u6587\u7ae0\u300125 \u7bc7\u5eb7\u5065\u96dc\u8a8c\u7684\u6587\u7ae0\u4ee5\u53ca\u91ab \u806f\u7db2 100 \u5247\u554f\u7b54\u505a\u6a19\u8a18\uff0c\u5f97\u5230 Fleiss' Kappa \u503c\u70ba 0.89\uff0c\u9054\u5230\u4e86 Landis \u4ee5\u53ca Koch \u6240\u8a8d\u70ba \u7684 Almost perfect agreement\uff0c\u78ba\u8a8d\u968e\u6bb5\u4e8c\u7684 Fleiss' Kappa \u6709\u660e\u986f\u4e0a\u5347\uff0c\u4e14\u9054\u5230\u53ef\u63a5\u53d7\u7684 \u7bc4\u570d\u5f8c\uff0c\u5269\u9918\u7684\u5f85\u6a19\u8a18\u7684\u4e2d\u6587\u53e5\u5b50\uff0c\u5247\u7531\u4e09\u4f4d\u6a19\u8a18\u4eba\u54e1\u5206\u5de5\u5404\u81ea\u6a19\u8a18\u3002 \u8868 1. \u547d\u540d\u5be6\u9ad4\u985e\u5225\u5b9a\u7fa9\u53ca\u7bc4\u4f8b \u985e\u5225 \u5b9a\u7fa9 \u7bc4\u4f8b \u4eba\u9ad4 (Body) \u6cdb\u6307\u751f\u7269\u9ad4\u7684\u7d30\u80de\u3001\u7d44\u7e54\u3001\u5668\u5b98\u548c\u7cfb\u7d71\u3002 \u7d30\u80de\u6838\u3001\u795e\u7d93\u7d44\u7e54\u3001\u5fc3\u3001 \u80ba\u3001\u810a\u9ad3\u3001\u547c\u5438\u7cfb\u7d71\u7b49\u3002 \u75c7\u72c0 (Symptom) \u53c8\u7a31\u75c5\u5fb5\uff0c\u7531\u60a3\u8005\u63cf\u8ff0\u7684\u4e3b\u89c0\u611f\u53d7\uff0c\u800c \u975e\u76f4\u63a5\u91cf\u6e2c\u5f97\u77e5\u3002 \u6d41\u9f3b\u6c34\u3001\u982d\u660f\u3001\u767c\u71d2\u3001 \u54b3\u55fd\u3001\u5931\u7720\u3001\u8ca7\u8840\u7b49\u3002 \u91ab\u7642\u5668\u6750 (Instrument) \u5305\u542b\u8a3a\u65b7\u3001\u6cbb\u7642\u3001\u6e1b\u8f15\u8207\u9810\u9632\u4eba\u985e\u75be\u75c5\uff0c \u4f7f\u7528\u7684\u5100\u5668\u3001\u5668\u68b0\u3001\u9644\u4ef6\u3001\u914d\u4ef6\u8207\u96f6\u4ef6\u3002 \u8840\u58d3\u8a08\u3001\u9054\u6587\u897f\u6a5f\u5668\u624b \u81c2\u3001\u4eba\u5de5\u9ad6\u95dc\u7bc0\u7b49\u3002 \u6aa2\u9a57 (Examination) \u5229\u7528\u91ab\u7642\u5668\u6750\u5c0d\u4eba\u9ad4\u5065\u5eb7\u72c0\u614b\u53ca\u751f\u7406\u529f \u80fd\u8a55\u4f30\u3002 \u807d \u529b \u6aa2 \u67e5 \u3001 \u986f \u5fae \u93e1 \u6aa2 \u67e5\u3001\u6838\u78c1\u5171\u632f\u9020\u5f71\u7b49\u3002 \u5316\u5b78\u7269\u8cea (Chemical) \u4eba\u9ad4\u7531\u4e0d\u540c\u7684\u5316\u5b78\u7269\u8cea\u7d44\u6210\uff0c\u96a8\u8457\u5e74\u9f61 \u8207\u5065\u5eb7\u72c0\u6cc1\u6709\u6240\u589e\u6e1b\u3002 \u53bb\u6c27\u6838\u7cd6\u6838\u9178\u3001\u4e09\u9178\u7518 \u6cb9\u916f\u3001\u7cd6\u5316\u8840\u8272\u7d20\u7b49\u3002 \u75be\u75c5 (Disease) \u6307\u4eba\u9ad4\u5728\u5916\u5728\u56e0\u7d20\u7684\u640d\u5bb3\u6216\u5167\u5728\u6a5f\u80fd\u4e0d \u826f\u60c5\u6cc1\u4e0b\uff0c\u5f71\u97ff\u90e8\u5206\u6216\u5168\u90e8\u5668\u5b98\u7570\u5e38\uff0c \u4f34\u96a8\u7279\u5b9a\u75c7\u72c0\u7684\u91ab\u5b78\u75c5\u75c7\u3002 \u5c0f\u5152\u9ebb\u75fa\u75c7\u3001\u5e15\u91d1\u68ee\u6c0f \u8166\u6ea2\u8840\u3001\u80ba\u7d50\u6838\u7b49\u3002 \u85e5\u54c1 (Drug) \u6cdb\u6307\u7528\u4f86\u505a\u8a3a\u65b7\u3001\u6cbb\u7642\u3001\u9810\u9632\u75be\u75c5\u6216\u6e1b \u8f15\u75db\u695a\u7684\u85e5\u7269\u6216\u5316\u5b78\u6210\u4efd\u3002 \u963f\u65af\u5339\u9748\u3001\u4e9e\u785d\u9178\u9209\u3001 \u4e9e\u9435\u9e7d\u3001\u6297\u751f\u7d20\u7b49 \u71df\u990a\u54c1 (Supplement) \u6307\u5f9e\u98df\u7269\u4e2d\u8403\u53d6\u5c0d\u4eba\u9ad4\u6709\u76ca\u7684\u71df\u990a\u7d20\uff0c \u4e3b\u8981\u529f\u80fd\u662f\u7dad\u6301\u5065\u5eb7\u548c\u9810\u9632\u75be\u75c5\u3002 \u81a0\u539f\u86cb\u767d\u3001\u76ca\u751f\u83cc\u3001\u7d9c \u5408\u7dad\u4ed6\u547d\u3001\u8449\u9ec3\u7d20\u7b49\u3002 \u6cbb\u7642 (Treatment) \u8b93\u60a3\u8005\u6062\u5fa9\u5065\u5eb7\u7684\u6cbb\u7652\u65b9\u5f0f\u3002 \u85e5\u7269\u6cbb\u7642\u3001\u8840\u6f3f\u7f6e\u63db\u3001 \u514d\u75ab\u7403\u86cb\u767d\u6ce8\u5c04\u7b49\u3002 \u6642\u9593 (Time) \u63cf\u8ff0\u60a3\u8005\u60a3\u75c5\u75c7\u72c0\u7684\u6301\u7e8c\u6642\u9593\u6216\u662f\u67d0\u500b \u6642\u523b\u3002 \u5b30\u5152\u671f\u3001\u5e7c\u5152\u6642\u671f\u3001\u9752 \u6625\u671f\u3001\u751f\u7406\u671f\u3001\u5b55\u671f\u7b49\u3002 \u76ee\u524d\u5728\u547d\u540d\u5be6\u9ad4\u8fa8\u8b58\u9818\u57df\u7684\u4e3b\u8981\u8a55\u4f30\u65b9\u6cd5\u70ba\u7cbe\u78ba\u7387 (Precision)\u3001\u53ec\u56de\u7387 (Recall)\u3001 \u5b8c\u5168\u76f8\u7b26\u624d\u7b97\u6b63\u78ba\u3002\u6df7\u6dc6\u77e9\u9663\u77e9\u9663\u7bc4\u4f8b\u5982\u8868 2\uff0c\u85c9\u6b64\u77e9\u9663\u8a08\u7b97\u7cbe\u78ba\u7387 (Precision) \u70ba\u300c\u6b63 \u78ba\u88ab\u8fa8\u8b58\u7684\u9805\u76ee\u300d\u5360\u300c\u7e3d\u8fa8\u8b58\u9805\u76ee\u300d\u7684\u6bd4\u4f8b\uff0c\u53ec\u56de\u7387 (Recall) \u70ba\u300c\u6b63\u78ba\u88ab\u8fa8\u8b58\u7684\u9805\u76ee\u300d \u986f\u8457\u512a\u65bc\u5148\u524d\u7684\u547d\u540d\u5be6\u9ad4\u8fa8\u8b58\u65b9\u6cd5\u3002 \u7a0b\u5f0f\u78bc\uff0c\u5c07\u8cc7\u6599\u66ff\u63db\u6210\u672c\u7814\u7a76\u6240\u4f7f\u7528\u7684\u8cc7\u6599\uff0c\u6a21\u578b\u8a2d\u5b9a\u7684\u53c3\u7167\u539f\u59cb\u7a0b\u5f0f\u78bc\uff0c\u800c\u6a21\u578b\u6703\u4f7f \u7d22\uff0c\u900f\u904e\u5716\u795e\u7d93\u7db2\u8def\u8abf\u9069\u81f3\u5065\u5eb7\u7167\u8b77\u547d\u540d\u5be6\u9ad4\u8fa8\u8b58\u4efb\u52d9\u3002\u6211\u5011\u7684\u6a21\u578b\u9054\u5230 75.69 \u7684 F1 \u6b64\u6a21\u578b\u70ba Zhang and Yang \u7b49\u4eba\u6240\u63d0\u51fa (Zhang & Yang, 2018)\uff0c\u5229\u7528\u5176\u8ad6\u6587\u4e2d\u63d0\u5230\u7684\u958b\u6e90 -radical -word 73.46 74.54 \u4e00\u3001\u6211\u5011\u63d0\u51fa\u4e00\u500b\u591a\u91cd\u5d4c\u5165\u5c0e\u5411\u7684\u5716\u5e8f\u5217\u7db2\u8def\u67b6\u69cb\uff0c\u5f9e\u90e8\u9996\u3001\u5b57\u5230\u8a5e\u7684\u4e0d\u540c\u8a9e\u610f\u8cc7\u8a0a\u88ab\u63a2 74.00 (4)\u3001Lattice (ACL 2018)\uff1a -word 73.48 75.10 \u6211\u5011\u63d0\u51fa\u9580\u63a7\u5716\u5e8f\u5217\u795e\u7d93\u7db2\u8def\u6a21\u578b\uff0c\u7528\u65bc\u4e2d\u6587\u5065\u5eb7\u7167\u8b77\u547d\u540d\u5be6\u9ad4\u8fa8\u8b58\u3002\u4e3b\u8981\u8ca2\u737b\u5982\u4e0b\uff1a 74.28 \u7dad\u7684\u5411\u91cf\u3002 -radical 73.50 76.73 75.08 5. \u7d50\u8ad6\u8207\u672a\u4f86\u7814\u7a76 (Conclusions and Future Work) \u6599\u5eab\u9032\u884c\u8a13\u7df4\u5373\u53ef\uff0c\u56e0\u6b64\u4f7f\u7528 4.2 \u7bc0\u4e2d\u6240\u63d0\u5230\u7684\u7dad\u57fa\u767e\u79d1\u7576\u4f5c\u8a13\u7df4\u8cc7\u6599\uff0c\u8a13\u7df4\u51fa\u5404 200 F1-score\uff0c\u5728\u672c\u7814\u7a76\u4e2d\u8a55\u4f30\u65b9\u5f0f\u63a1\u7cbe\u6e96\u6bd4\u5c0d (exact match)\uff0c\u610f\u5373\u9810\u6e2c\u7684\u7d50\u679c\u9700\u8207\u6b63\u78ba\u7d50\u679c \u6b64\u5c07\u8cc7\u6599\u66ff\u63db\u6210\u672c\u7814\u7a76\u6240\u4f7f\u7528\u7684\u8cc7\u6599\uff0c\u53c3\u6578\u7684\u8a2d\u5b9a\u8207\u539f\u59cb\u7a0b\u5f0f\u78bc\u76f8\u540c\uff0c\u7531\u65bc\u958b\u6e90\u7a0b\u5f0f\u78bc GGSNN (ours) 75.46 75.76 75.69 \u4e26\u672a\u63d0\u4f9b\u6a21\u578b\u6240\u6703\u7528\u5230\u7684\u7684\u5b57\u5d4c\u5165\u4ee5\u53ca\u4e8c\u5143\u5d4c\u5165\uff0c\u800c\u5728\u5176\u5b98\u7db2\u7684\u8aaa\u660e\u70ba\u4f7f\u7528\u7dad\u57fa\u767e\u79d1\u8a9e Lattice (ACL 2018) 74.69 75.76 \u9810\u6e2c \u9240\u96e2\u5b50\u91cf\u82e5\u651d\u53d6\u5145\u8db3\uff0c\u53ef\u964d\u4f4e\u8166\u8840\u7ba1 \u963b\u585e \u98a8\u96aa\u3002 75.22 NO-CROSS 4.4 \u6548\u80fd\u8a55\u4f30 (Evaluation) \u6b64\u6a21\u578b\u70ba Ding \u7b49\u4eba\u6240\u63d0\u51fa (Ding et al., 2019)\uff0c\u5728\u5176\u767c\u8868\u7684\u8ad6\u6587\u4e2d\u6709\u63d0\u4f9b\u958b\u6e90\u7a0b\u5f0f\u78bc\uff0c\u56e0 73.00 75.56 74.26 \u7b54\u6848 \u9240\u96e2\u5b50\u91cf\u82e5\u651d\u53d6\u5145\u8db3\uff0c\u53ef\u964d\u4f4e\u8166\u8840\u7ba1 \u963b\u585e \u98a8\u96aa\u3002 Gazetteers (ACL 2019) \u75c7\u3001\u6182\u9b31\u75c7\u3001\u9752\u5149\u773c\u3001 \u5b57 (29.99 \u500b\u8a5e)\uff0c\u7e3d\u5171\u6709 61,155 \u500b\u547d\u540d\u5be6\u9ad4\uff0c\u5e73\u5747\u6bcf\u500b\u53e5\u5b50\u6709 2.17 \u500b\u3002\u6e2c\u8a66\u8cc7\u6599\u4f86\u81ea\u4e09 \u500b\u6a19\u8a18\u4eba\u54e1\u5171\u540c\u6a19\u8a18\u6709\u4e00\u81f4\u7d50\u679c\u7684 2,531 \u53e5\uff0c\u6bcf\u500b\u53e5\u5b50\u5e73\u5747 47.92 \u500b\u5b57 (28.67 \u500b\u8a5e)\uff0c\u7e3d \u8868 2. \u6df7\u6dc6\u77e9\u9663 \uf0b7 CONTAIN\uff1a\u6b63\u78ba\u7684\u547d\u540d\u5be6\u9ad4\u300c\u5305\u542b\u300d\u9810\u6e2c\u7684\u547d\u540d\u5be6\u9ad4\u3002 \u72c0\u3001\u91ab\u7642\u5668\u6750\u3001\u6aa2\u9a57\u3001\u5316\u5b78\u7269\u8cea\u3001\u75be\u75c5\u3001\u85e5\u54c1\u3001\u71df\u990a\u54c1\u3001\u6cbb\u7642\u4ee5\u53ca\u6642\u9593\u3002 \u6b64\u70ba\u672c\u7814\u7a76\u63d0\u51fa\u7684\u6a21\u578b\uff0c\u5728\u7b2c\u4e09\u7ae0\u6709\u8a73\u7d30\u7684\u4ecb\u7d39\u3002\u5176\u4e2d -radical \u70ba\u6b64\u6a21\u578b GGSNN \u53bb\u9664 [Table 2. The confusion matrix] \u90e8\u9996\u5d4c\u5165\u3002-word \u70ba\u6b64\u6a21\u578b GGSNN \u53bb\u9664\u8a5e\u5d4c\u5165\u3002 -radical -word \u5247\u70ba\u6b64\u6a21\u578b GGSNN \u5229\u7528\u547d\u540d\u5be6\u9ad4\u8fa8\u8b58\u7684\u9019\u9805\u6280\u8853\uff0c\u6211\u5011\u53ef\u4ee5\u4f9d\u7167\u5404\u9818\u57df\u4e0d\u540c\u7684\u9700\u6c42\uff0c\u5f9e\u975e\u7d50\u69cb\u7684\u6587\u7ae0 \uf0b7 CONTAINED\uff1a\u6b63\u78ba\u7684\u547d\u540d\u5be6\u9ad4\u300c\u88ab\u5305\u542b\u65bc\u300d\u9810\u6e2c\u7684\u547d\u540d\u5be6\u9ad4\u3002 \u5171\u6709 7,305 \u500b\u547d\u540d\u5be6\u9ad4\uff0c\u5e73\u5747\u6bcf\u500b\u53e5\u5b50\u6709 2.89 \u500b\u300210 \u500b\u985e\u5225\u5728\u8a13\u7df4\u548c\u6e2c\u8a66\u8cc7\u6599\u5206\u4f48\u76f8\u4f3c\uff0c \u6700\u591a\u7684\u547d\u540d\u5be6\u9ad4\u985e\u5225\u662f\u4eba\u9ad4\uff0c\u7d04\u4f54 38%\uff0c\u4f9d\u5e8f\u662f\u75c7\u72c0\u3001\u75be\u75c5\u548c\u5316\u5b78\u7269\u8cea\uff0c\u524d\u56db\u5927\u985e\u4f54\u7e3d \u6578\u7684 82%\uff0c\u5176\u9918 6 \u985e\u7d04\u4f54\u7e3d\u6578\u7684 18\uff05\u3002 4.2 \u5d4c\u5165\u5411\u91cf (Embedding) \u672c\u7814\u7a76\u6240\u4f7f\u7528\u7684\u5d4c\u5165\u65b9\u5f0f\u70ba Word2vec\uff0c\u8a13\u7df4\u7684\u8cc7\u6599\u4f86\u6e90\u70ba\u7dad\u57fa\u767e\u79d1\uff0c\u4e0b\u8f09\u8a9e\u6599\u5eab\u7684\u65e5\u671f \u70ba 2020 \u5e74 2 \u6708 3 \u65e5\uff0c\u5229\u7528\u6b64\u6a94\u6848\u6211\u5011\u53ef\u4ee5\u8a13\u7df4\u51fa\u5b57\u5d4c\u5165\u3001\u90e8\u9996\u5d4c\u5165\u4ee5\u53ca\u8a5e\u5d4c\u5165\uff0c\u8a5e\u983b\u8a2d \u5b9a\u70ba\u81f3\u5c11\u51fa\u73fe 5 \u6b21\u4ee5\u4e0a\uff0c\u5411\u91cf\u7684\u7dad\u5ea6\u7684\u8a2d\u5b9a\u7686\u70ba 50 \u7dad\uff0c\u6700\u7d42\u7372\u5f97 863,835 \u500b\u8a5e\u5d4c\u5165\u5411\u91cf\uff0c 13,581 \u500b\u5b57\u5d4c\u5165\u5411\u91cf\u4ee5\u53ca 3,209 \u500b\u90e8\u9996\u5d4c\u5165\u5411\u91cf\u3002 4.3 \u5be6\u9a57\u8a2d\u5b9a (Settings) \u672c\u7814\u7a76\u6240\u4f7f\u7528\u7684\u5b57\u5178\u4f86\u6e90\u4e00\u5171\u5206\u70ba\u4e09\u500b\uff0c\u5206\u5225\u70ba\u570b\u5bb6\u7db2\u8def\u91ab\u85e5 4 \u3001\u570b\u5bb6\u6559\u80b2\u7814\u7a76\u9662 5 \u4ee5\u53ca \u641c\u72d7\u7db2 6 \uff0c\u5176\u4e2d\u570b\u5bb6\u7db2\u8def\u91ab\u85e5\u7684\u8a5e\u5f59\u4e3b\u8981\u70ba\u5e38\u898b\u7684\u91ab\u8b77\u540d\u8a5e\uff0c\u570b\u5bb6\u6559\u80b2\u7814\u7a76\u9662\u9078\u7528\u7684\u8cc7\u6599 \u70ba\u91ab\u5b78\u540d\u8a5e\uff0c\u800c\u641c\u72d7\u7db2\u6240\u5305\u542b\u7684\u5167\u5bb9\u70ba ICD-10\u3001\u4eba\u9ad4\u7a74\u4f4d\u540d\u7a31\u3001\u91ab\u5b78\u8a5e\u5f59\u3001\u91ab\u7642\u6aa2\u9a57\u4ee5 \u53ca\u91ab\u7642\u5668\u6750\u7b49\u7b49\uff0c\u5728\u4f7f\u7528\u5b57\u5178\u6642\uff0c\u5c07\u4e0a\u8ff0\u5b57\u5178\u5148\u5408\u4f75\u5f8c\u5206\u985e\uff0c\u4f9d\u7167\u8a5e\u5f59\u5b57\u6578\u4e00\u5171\u5206\u6210\u4e94 \u500b\u5b57\u5178\uff0c1 \u500b\u5b57\u7684\u5b57\u5178\u6709 351 \u500b\u8a5e\uff0c2 \u500b\u5b57\u7684\u5b57\u5178\u6709 7,978 \u500b\u8a5e\uff0c3 \u500b\u5b57\u7684\u5b57\u5178\u6709 19,282 \u500b\u8a5e\uff0c4 \u500b\u5b57\u7684\u5b57\u5178\u6709 31,444 \u500b\u8a5e\uff0c\u8a5e\u5f59\u5b57\u6578\u70ba 5 \u500b\u5b57\u4ee5\u4e0a\u7684\u5b57\u5178\u6709 95,362 \u500b\u8a5e\u3002 \u5728\u8a13\u7df4\u904e\u7a0b\u4e2d\u5b78\u7fd2\u7387 (learning rate) \u4ee5\u53ca\u8a13\u7df4\u8cc7\u6599\u6703\u96a8\u8457\u6642\u671f (epoch) \u8abf\u6574\uff0c\u55ae\u6578 epoch \u7684 learning rate \u70ba 0.001\uff0c\u8cc7\u6599\u70ba\u539f\u59cb\u6574\u4efd\u7684\u8a13\u7df4\u8cc7\u6599\uff0c\u96d9\u6578 epoch \u7684 learning rate \u70ba 0.0005\uff0c\u8cc7\u6599\u70ba\u5c1a\u672a\u5b78\u7fd2\u597d\u7684\u8a13\u7df4\u8cc7\u6599\uff0c\u5224\u65b7\u7684\u4f9d\u64da\u70ba\u547d\u540d\u5be6\u9ad4\u8fa8\u8b58\u662f\u5426\u6709\u932f\u8aa4\u3002\u5176 \u4e2d\u4e4b\u6240\u4ee5\u6703\u91dd\u5c0d\u5c1a\u672a\u5b78\u7fd2\u597d\u7684\u8cc7\u6599\u518d\u5b78\u7fd2\u4e00\u904d\u7684\u539f\u56e0\u70ba\u7406\u8ad6\u4e0a\u5728\u8a13\u7df4\u7684\u904e\u7a0b\u4e2d\uff0c\u6703\u5e0c\u671b \u6a21\u578b\u80fd\u5920\u5c07\u6240\u6709\u7684\u8a13\u7df4\u8cc7\u6599\u5b78\u7fd2\u6b63\u78ba\uff0cepoch \u7684\u8a2d\u5b9a\u503c\u70ba 80\uff0c batch size \u70ba 32\uff0cLSTM \u96b1 (3)\u3001Gazetteers (ACL 2019)\uff1a ME-CNER (CIKM 2019) 73.68 74.62 74.15 \u9810\u6e2c \u5c0d\u65bc \u75f0\u6fc1\u7600 \u963b\u7d93\u7d61 \u800c\u81f4\u7684\u75c7\u72c0\u6709\u6539\u5584\u7684\u529f\u80fd\u3002 \u85cf\u5c64\u7684\u7dad\u5ea6\u70ba 200 \u7dad\uff0cGGSNN \u7684\u66f4\u65b0\u6b21\u6578 (time step) \u8a2d\u5b9a\u70ba 2\u3002 \u771f\u5be6\u503c \u9810\u6e2c\u503c Positive Negative Positive True Positive (TP) False Negative (FN) Negative False Positive (FP) True Negative (TF) Pr ecision \uf03d TP TP \uf02b FP (15) Re call \uf03d TP TP \uf02b FN (16) F1\uf02d score \uf03d 2 * Pr ecision * Re call Pr ecision \uf02b Re call (17) 4.5 \u5be6\u9a57\u7d50\u679c (Results) \u6211\u5011\u6bd4\u8f03\u4e86\u4ee5\u4e0b\u4e2d\u6587\u547d\u540d\u5be6\u9ad4\u6a21\u578b\u7684\u6548\u80fd\u5dee\u7570 (1)\u3001BiLSTM-CRF (ICCPOL 2016)\uff1a \u6b64\u6a21\u578b\u5be6\u4f5c\u4e86 Dong \u7b49\u4eba(2016) \u7684\u67b6\u69cb\uff0c\u4ee5\u5b57\u4f5c\u70ba\u57fa\u790e\u7576\u4f5c\u6a21\u578b\u8f38\u5165\uff0c\u5b57\u5d4c\u5165\u4f7f\u7528\u900f\u904e 4.2 \u7bc0\u4e2d\u6240\u63d0\u5230\u7684\u7dad\u57fa\u767e\u79d1\u8a9e\u6599\u5eab\u7576\u4f5c\u8a13\u7df4\u8cc7\u6599\uff0c\u5411\u91cf\u7dad\u5ea6\u70ba 200 \u7dad\u3002 (2)\u3001ME-CNER (CIKM 2019)\uff1a Xu \u7b49\u4eba(2019) \u63d0\u51fa\u7684\u6a21\u578b\uff0c\u672c\u7814\u7a76\u5be6\u4f5c\u7684\u6a21\u578b\u67b6\u69cb\u5c07\u5176\u7a0d\u505a\u66f4\u52d5\uff0c\u672c\u7814\u7a76\u8a8d\u70ba\u5c07\u5b57\u5d4c Convolutions \u5f8c\u9023\u63a5\uff0c\u5176\u4e2d\u524d\u8005\u7684 BiLSTM \u8f03\u80fd\u4fdd\u7559\u539f\u59cb\u7684\u8a0a\u606f\uff0c\u4e14\u6bd4\u539f\u6a21\u578b\u6548\u80fd\u597d\u3002 BiLSTM-CRF (ICCPOL 2016) 70.38 72.77 CROSS 71.56 \u7b54\u6848 \u5c0d\u65bc \u75f0\u6fc1 \u7600\u963b\u7d93\u7d61 \u800c\u81f4\u7684\u75c7\u72c0\u6709\u6539\u5584\u7684\u529f\u80fd\u3002 \u5165 \u5206 \u5225 \u7d93 \u904e BiLSTM \u4ee5 \u53ca Convolutions \u6bd4 \u8d77 \u5206 \u5225 \u7d93 \u904e BiLSTM-Convolution \u4ee5 \u53ca \u540c\u6642\u53bb\u9664\u90e8\u9996\u5d4c\u5165\u4ee5\u53ca\u8a5e\u5d4c\u5165\u3002 \u8868 3 \u70ba\u6a21\u578b\u6548\u80fd\u6bd4\u8f03\u7d50\u679c\u3002ME-CNER \u8207 BiLSTM-CRF \u5169\u8005\u7684\u5dee\u7570\u70ba\u662f\u5426\u6709\u52a0\u5165\u90e8 \u9996\u5d4c\u5165\u4ee5\u53ca\u8a5e\u5d4c\u5165\uff0c\u5f9e\u5be6\u9a57\u7d50\u679c\u5f97\u77e5 ME-CNER \u76f8\u8f03\u65bc BiLSTM-CRF \u63d0\u5347\u4e86 2.59 \u7684 F1\uff0c \u4e2d\u62bd\u53d6\u51fa\u8a72\u9818\u57df\u6240\u95dc\u6ce8\u7684\u547d\u540d\u5be6\u9ad4\uff0c\u900f\u904e\u9019\u4e9b\u62bd\u53d6\u51fa\u7684\u547d\u540d\u5be6\u9ad4\uff0c\u6211\u5011\u53ef\u4ee5\u5145\u5206\u7684\u638c\u63e1 \uf0b7 SPLIT\uff1a\u6b63\u78ba\u7684\u547d\u540d\u5be6\u9ad4\u6216\u662f\u9810\u6e2c\u7684\u547d\u540d\u5be6\u9ad4\u88ab\u62c6\u6210\u5169\u6bb5\u547d\u540d\u5be6\u9ad4\u3002 \u6587\u7ae0\u4e2d\u7684\u8cc7\u8a0a\uff0c\u5c0d\u6587\u7ae0\u505a\u66f4\u9032\u4e00\u6b65\u7684\u5206\u6790\uff0c\u5728\u672a\u4f86\u7684\u61c9\u7528\u4e2d\uff0c\u547d\u540d\u5be6\u9ad4\u8fa8\u8b58\u6240\u6a19\u793a\u51fa\u7684 \uf0b7 CROSS\uff1a\u6b63\u78ba\u7684\u547d\u540d\u5be6\u8207\u9810\u6e2c\u7684\u547d\u540d\u5be6\u9ad4\u4e4b\u9593\u300c\u6709\u300d\u91cd\u758a\u7684\u5b57\u3002 \u547d\u540d\u5be6\u9ad4\uff0c\u53ef\u4ee5\u505a\u70ba\u95dc\u4fc2\u64f7\u53d6\u3001\u4e8b\u4ef6\u5075\u6e2c\u8207\u8ffd\u8e64\u3001\u77e5\u8b58\u5716\u8b5c\u5efa\u7f6e\u3001\u667a\u6167\u554f\u7b54\u7cfb\u7d71\u7b49\u61c9\u7528 \uf0b7 NO-CROSS\uff1a\u6b63\u78ba\u7684\u547d\u540d\u5be6\u9ad4\u8207\u9810\u6e2c\u7684\u547d\u540d\u5be6\u9ad4\u4e4b\u9593\u300c\u6c92\u6709\u300d\u91cd\u758a\u7684\u5b57\u3002 \u7684\u57fa\u790e\u3002 \u56e0\u6b64\u52a0\u5165\u90e8\u9996\u5d4c\u5165\u4ee5\u53ca\u8a5e\u5d4c\u5165\u6709\u52a9\u65bc\u63d0\u5347\u6a21\u578b\u7684\u8868\u73fe\u3002Gazetteers \u8207 BiLSTM-CRF \u5169\u8005 5 \u7a2e\u985e\u578b\u7684\u932f\u8aa4\u7e3d\u5171\u6709 2,193 \u500b\uff0c\u5176\u4e2d\u6700\u591a\u7684\u932f\u8aa4\u985e\u578b\u70ba NO-CROSS\uff0c\u7d04\u4f54 72%\u3002 \u4e3b\u8981\u7684\u5dee\u7570\u70ba\u662f\u5426\u52a0\u5165\u5b57\u5178\u7684\u8cc7\u8a0a\uff0c\u5f9e\u5be6\u9a57\u7d50\u679c\u5f97\u77e5 Gazetteers \u76f8\u8f03\u65bc BiLSTM-CRF \u63d0 Precision Recall F1 \u5019\u7fa4 (BPH) \u548c\u80de\u6f3f\u7cbe\u5b50\u6ce8\u5c04 (ICSI) \u7b49\uff0c\u6c92\u6709\u5728\u8a13\u7df4\u8cc7\u6599\u4e2d\uff0c\u4e5f\u4e0d\u5c6c\u65bc\u5b57\u5178\u4e2d\u7684\u8a5e\u5f59\uff0c \u7121\u6cd5\u88ab\u6b63\u78ba\u8fa8\u8b58\u3002\u85c9\u7531\u932f\u8aa4\u5206\u6790\u5f97\u77e5\uff0c\u5b57\u5178\u8a5e\u5f59\u6db5\u84cb\u7a0b\u5ea6\u5c0d\u6a21\u578b\u6548\u80fd\u6709\u91cd\u8981\u7684\u5f71\u97ff\u3002 \u8868 4. \u547d\u540d\u5be6\u9ad4\u9810\u6e2c\u932f\u8aa4\u985e\u578b\u8207\u7bc4\u4f8b CONTAIN \u7b54\u6848 \u570b\u969b\u9593 \u5fb7\u570b\u9ebb\u75b9 \u4ecd\u6709\u75ab\u60c5\u767c\u751f\uff0c\u6240\u4ee5\u6709\u51fa\u570b\u8a08\u756b\u8981\u9810 \u5148\u505a\u597d\u5b89\u6392\u3002 \u9810\u6e2c \u570b\u969b\u9593 \u5fb7\u570b\u9ebb\u75b9 \u4ecd\u6709\u75ab\u60c5\u767c\u751f\uff0c\u6240\u4ee5\u6709\u51fa\u570b\u8a08\u756b\u8981\u9810 \u5148\u505a\u597d\u5b89\u6392\u3002 CONTAINED \u7b54\u6848 \u80ba\u4e3b\u8108 \u6307\u6a6b\u8188\u819c \u929c\u63a5\u5fc3\u81df\u7684\u90e8\u5206\u3002 \u9810\u6e2c \u80ba\u4e3b\u8108 \u6307\u6a6b\u8188\u819c \u929c\u63a5\u5fc3\u81df\u7684\u90e8\u5206\u3002 SPLIT \u7b54\u6848 \u5589\u56a8\u75db \u4e3b\u8981\u662f\u6211\u5011\u7684\u6241\u6843\u817a\u767c\u708e\u3002 \u9810\u6e2c \u5589\u56a8 \u75db \u4e3b\u8981\u662f\u6211\u5011\u7684\u6241\u6843\u817a\u767c\u708e\u3002 \u5347\u4e86 2.7 \u7684 Method \u6211\u5011\u89c0\u5bdf\u5f8c\u5f97\u77e5\uff0c\u6709\u4e9b\u9818\u57df\u8a5e\u5f59\u4f8b\u5982\uff1a\u8840\u6e05\u80fa\u57fa\u4e19\u916e\u9178\u8f49\u5316\u9176 (SGPT)\u3001\u651d\u8b77\u817a\u80a5\u5927\u75c7 \u81f4\u8b1d (Acknowledgements)</td></tr></table>" | |
| } | |
| } | |
| } | |
| } |