| { | |
| "paper_id": "2020", | |
| "header": { | |
| "generated_with": "S2ORC 1.0.0", | |
| "date_generated": "2023-01-19T12:54:00.050950Z" | |
| }, | |
| "title": "\u57fa \u57fa \u57fa\u4e8e \u4e8e \u4e8eBERT\u7684 \u7684 \u7684\u7aef \u7aef \u7aef\u5230 \u5230 \u5230\u7aef \u7aef \u7aef\u4e2d \u4e2d \u4e2d\u6587 \u6587 \u6587\u7bc7 \u7bc7 \u7bc7\u7ae0 \u7ae0 \u7ae0\u4e8b \u4e8b \u4e8b\u4ef6 \u4ef6 \u4ef6\u62bd \u62bd \u62bd\u53d6 \u53d6 \u53d6", | |
| "authors": [ | |
| { | |
| "first": "\u5f20", | |
| "middle": [ | |
| "\u5f20" | |
| ], | |
| "last": "\u5f20\u6d2a", | |
| "suffix": "", | |
| "affiliation": { | |
| "laboratory": "", | |
| "institution": "Donghua University", | |
| "location": { | |
| "postCode": "201620", | |
| "settlement": "Shanghai", | |
| "country": "China" | |
| } | |
| }, | |
| "email": "" | |
| }, | |
| { | |
| "first": "\u6d2a", | |
| "middle": [], | |
| "last": "\u6d2a\u5bbd \u5bbd \u5bbd \u5b8b \u5b8b \u5b8b\u6656 \u6656 \u6656 ( ) \u738b \u738b \u738b\u8212 \u8212 \u8212\u6021 \u6021 \u6021 \u5f90 \u5f90 \u5f90\u6ce2", | |
| "suffix": "", | |
| "affiliation": { | |
| "laboratory": "", | |
| "institution": "Donghua University", | |
| "location": { | |
| "postCode": "201620", | |
| "settlement": "Shanghai", | |
| "country": "China" | |
| } | |
| }, | |
| "email": "" | |
| }, | |
| { | |
| "first": "Hongkuan", | |
| "middle": [], | |
| "last": "Zhang", | |
| "suffix": "", | |
| "affiliation": { | |
| "laboratory": "", | |
| "institution": "Donghua University", | |
| "location": { | |
| "postCode": "201620", | |
| "settlement": "Shanghai", | |
| "country": "China" | |
| } | |
| }, | |
| "email": "" | |
| }, | |
| { | |
| "first": "Hui", | |
| "middle": [], | |
| "last": "Song", | |
| "suffix": "", | |
| "affiliation": { | |
| "laboratory": "", | |
| "institution": "Donghua University", | |
| "location": { | |
| "postCode": "201620", | |
| "settlement": "Shanghai", | |
| "country": "China" | |
| } | |
| }, | |
| "email": "[email protected]" | |
| }, | |
| { | |
| "first": "Shuyi", | |
| "middle": [], | |
| "last": "Wang", | |
| "suffix": "", | |
| "affiliation": { | |
| "laboratory": "", | |
| "institution": "Donghua University", | |
| "location": { | |
| "postCode": "201620", | |
| "settlement": "Shanghai", | |
| "country": "China" | |
| } | |
| }, | |
| "email": "" | |
| }, | |
| { | |
| "first": "Bo", | |
| "middle": [], | |
| "last": "Xu", | |
| "suffix": "", | |
| "affiliation": { | |
| "laboratory": "", | |
| "institution": "Donghua University", | |
| "location": { | |
| "postCode": "201620", | |
| "settlement": "Shanghai", | |
| "country": "China" | |
| } | |
| }, | |
| "email": "[email protected]" | |
| } | |
| ], | |
| "year": "", | |
| "venue": null, | |
| "identifiers": {}, | |
| "abstract": "Document-level event extraction aims at discovering event mentions and extracting events which contain event arguments and their roles from texts. This paper proposes an end-to-end model for closed-domain based on BERT. We introduce the embedding of event type and entity nodes to the subsequent layer for event argument and role identification, which represents the relation between event, arguments and roles and improves the accuracy of classifying multi-event arguments. With the title, the quintuple of event, we calculate the master slave structure between multiple events with the embedding presentation. Experimental results show that our model outperforms the state of the art.", | |
| "pdf_parse": { | |
| "paper_id": "2020", | |
| "_pdf_hash": "", | |
| "abstract": [ | |
| { | |
| "text": "Document-level event extraction aims at discovering event mentions and extracting events which contain event arguments and their roles from texts. This paper proposes an end-to-end model for closed-domain based on BERT. We introduce the embedding of event type and entity nodes to the subsequent layer for event argument and role identification, which represents the relation between event, arguments and roles and improves the accuracy of classifying multi-event arguments. With the title, the quintuple of event, we calculate the master slave structure between multiple events with the embedding presentation. Experimental results show that our model outperforms the state of the art.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Abstract", | |
| "sec_num": null | |
| } | |
| ], | |
| "body_text": [ | |
| { | |
| "text": ".\u6587\u6863\u4e2d\u5b58\u5728\"\u4e34\u65f6\u505c\u724c\"\u4e0e\"\u590d\u724c\"\u4e24\u4e2a\u4e8b\u4ef6\uff0c\u4e14\u5171\u4eab\u4e8b\u4ef6\u5143\u7d20\"N\u534e\u76db\u660c(\u89d2\u8272\u4e3a\u8bc1\u5238\u7b80 \u79f0)\"\u548c\"\u672c\u6240(\u89d2\u8272\u4e3a\u6267\u884c\u65b9)\"\u3002", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "\u56fe1", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "\u76ee\u524d\u7bc7\u7ae0\u7ea7\u7684\u4e8b\u4ef6\u62bd\u53d6\u5927\u591a\u91c7\u7528\u7ba1\u9053\u5f0f\u65b9\u6cd5\uff0c\u5373\u5206\u522b\u5bf9\u4e8b\u4ef6\u548c\u4e8b\u4ef6\u5143\u7d20\u8fdb\u884c\u8bc6\u522b\uff0c\u7136\u540e\u5224 \u65ad\u5143\u7d20\u7684\u89d2\u8272\u3002\u5982Yang (2018)\u57fa\u4e8e\u53e5\u5b50\u62bd\u53d6\u7ed3\u679c\u53ca\u6587\u672c\u7279\u5f81\u53d1\u73b0\u6838\u5fc3\u4e8b\u4ef6\uff0c\u91c7\u7528\u5143\u7d20\u8865\u9f50\u7b56\u7565 \u5f97\u5230\u7bc7\u7ae0\u7ea7\u522b\u7684\u4e8b\u4ef6\u4fe1\u606f\u3002ZHONG (2019)\u5728\u53e5\u5b50\u7ea7\u62bd\u53d6\u7ed3\u679c\u7684\u57fa\u7840\u4e0a\u5229\u7528\u6574\u6570\u7ebf\u6027\u89c4\u5212\u8fdb\u884c \u5168\u5c40\u63a8\u7406\uff0c\u878d\u5408\u5171\u6307\u4e8b\u4ef6\u7684\u5143\u7d20\uff0c\u5b9e\u73b0\u7bc7\u7ae0\u7ea7\u4e8b\u4ef6\u62bd\u53d6\u3002\u7531\u4e8e\u7ba1\u9053\u6a21\u578b\u672c\u8eab\u7684\u5c40\u9650\u6027\uff0c\u5bfc\u81f4\u8fd9 \u4e9b\u65b9\u6cd5\u6574\u4f53\u6548\u679c\u4e0d\u7406\u60f3\u3002Yang (2016)\u63d0\u51fa\u4e00\u4e2a\u7aef\u5230\u7aef\u5b66\u4e60\u6a21\u578b\uff0c\u901a\u8fc7\u91c7\u7528\u8054\u5408\u56e0\u5b50\u56fe\u6a21\u578b\u6765\u8054 \u5408\u5b66\u4e60\u6bcf\u4e2a\u4e8b\u4ef6\u5185\u90e8\u7684\u7ed3\u6784\u5316\u4fe1\u606f\u3001\u7bc7\u7ae0\u5185\u4e0d\u540c\u4e8b\u4ef6\u95f4\u7684\u5173\u7cfb\u548c\u5b9e\u4f53\u4fe1\u606f\uff0c\u4f46\u662f\u8be5\u5de5\u4f5c\u4f7f\u7528\u4e86 \u5927\u91cf\u4eba\u5de5\u7279\u5f81\uff0c\u4e0d\u5229\u4e8e\u9886\u57df\u6cdb\u5316\u3002 Zheng (2019)\u91c7\u7528\u7aef\u5230\u7aef\u6a21\u578b\u9996\u5148\u8bc6\u522b\u53e5\u5b50\u7ea7\u522b\u7684\u4e8b\u4ef6\u5143\u7d20\uff0c\u518d\u5229\u7528\u4e8c\u5206\u7c7b\u7b56\u7565\u5224\u5b9a\u4e8b\u4ef6 \u7c7b\u578b\uff0c\u7136\u540e\u5c06\u5143\u7d20\u89d2\u8272\u8bc6\u522b\u8f6c\u5316\u4e3a\u6709\u5411\u65e0\u73af\u56fe\u7684\u751f\u6210\u8fc7\u7a0b\u3002\u8be5\u6587\u80fd\u591f\u8f83\u597d\u5730\u5904\u7406\u5355\u4e8b\u4ef6\u5355\u5b9e\u4f8b \u548c\u5355\u4e8b\u4ef6\u591a\u5b9e\u4f8b\u7684\u6837\u672c\uff0c\u4f46\u5728\u5143\u7d20\u5206\u5e03\u8f83\u5206\u6563\u4e14\u5b58\u5728\u5143\u7d20\u91cd\u5408\u7684\u591a\u4e8b\u4ef6\u591a\u5b9e\u4f8b\u6837\u672c\u4e0a\u6027\u80fd\u7565 \u5dee\u3002\u9488\u5bf9\u4ee5\u4e0a\u95ee\u9898\uff0c\u672c\u6587\u5728\u5df2\u6709\u7bc7\u7ae0\u7ea7\u7684\u62bd\u53d6\u7814\u7a76\u5de5\u4f5c\u57fa\u7840\u4e0a\uff0c\u63d0\u51fa\u4e86\u57fa\u4e8eBERT (Devlin al et., 2019)\u7684\u7aef\u5230\u7aef\u4e8b\u4ef6\u62bd\u53d6\u6a21\u578bDLEMC(Document-Level End-to-end Model in Chinese)\uff0c \u8be5\u6a21\u578b\u65e0\u9700\u5c06\u6587\u7ae0\u6309\u53e5\u5b50\u5206\u5272\u5206\u522b\u5904\u7406\uff0c\u5c3d\u53ef\u80fd\u4fdd\u7559\u5b8c\u6574\u7684\u6587\u672c\uff0c\u51cf\u5c11\u4fe1\u606f\u635f\u5931\uff1b\u5728\u4e8b\u4ef6\u5143\u7d20 \u8bc6\u522b\u4e2d\u5f15\u5165\u4e86\u4e8b\u4ef6\u7c7b\u578b\u5d4c\u5165\u589e\u5f3a\u6587\u672c\u7279\u5f81\uff0c\u5e76\u5728\u5143\u7d20\u89d2\u8272\u5206\u7c7b\u4e2d\u5229\u7528\u6ce8\u610f\u529b\u673a\u5236\u5f15\u5165\u4e8b\u4ef6\u7c7b\u578b \u548c\u5b9e\u4f53\u5d4c\u5165\u8868\u793a\uff0c\u4ee5\u4fbf\u51c6\u786e\u8bc6\u522b\u4e8b\u4ef6\u5143\u7d20\u53ca\u5176\u5728\u4e0d\u540c\u4e8b\u4ef6\u4e2d\u7684\u89d2\u8272\u3002 DLEMC\u6a21\u578b\u5206\u4e3a4\u5c42\uff1a\u8f93\u5165\u7f16\u7801\u5c42\u3001\u4e8b\u4ef6\u68c0\u6d4b\u5c42\u3001\u4e8b\u4ef6\u5143\u7d20\u8bc6\u522b\u5c42\u4ee5\u53ca\u5143\u7d20\u89d2\u8272\u8bc6\u522b\u5c42\u3002 \u8f93\u5165\u7f16\u7801\u5c42\u63a5\u6536\u8f93\u5165\u7684\u6587\u672c\u4fe1\u606f\u5e76\u8f93\u51fa\u5bf9\u5e94\u7684\u5d4c\u5165\u8868\u793a\uff1b\u4e8b\u4ef6\u68c0\u6d4b\u5c42\u91c7\u7528\u591a\u4e2a\u5206\u7c7b\u5668\u5bf9\u540c\u4e00\u4e2a \u6587\u672c\u7279\u5f81\u5411\u91cf\u8fdb\u884c\u591a\u6807\u7b7e\u4e8b\u4ef6\u5206\u7c7b\uff1b\u4e8b\u4ef6\u5143\u7d20\u8bc6\u522b\u5c42\u901a\u8fc7\u5f15\u5165\u4e8b\u4ef6\u7279\u5f81\u5b66\u4e60\u4e0d\u540c\u4e8b\u4ef6\u4e2d\u5143\u7d20\u7684 \u8bed\u4e49\u4fe1\u606f\uff0c\u8fdb\u884c\u4e8b\u4ef6\u5143\u7d20\u7684\u8bc6\u522b\uff1b\u5143\u7d20\u89d2\u8272\u8bc6\u522b\u5c42\u5229\u7528\u6ce8\u610f\u529b\u673a\u5236\u6765\u63d0\u9ad8\u6a21\u578b\u5bf9\u786e\u5b9a\u4e8b\u4ef6\u4e2d\u6bcf \u4e2a\u5143\u7d20\u7279\u5f81\u7684\u5173\u6ce8\u5ea6\uff0c\u5224\u65ad\u5176\u5728\u5bf9\u5e94\u4e8b\u4ef6\u4e2d\u626e\u6f14\u7684\u89d2\u8272\u3002\u6700\u540e\u57fa\u4e8e\u4e8b\u4ef6\u7684\u5d4c\u5165\u8868\u793a\u8ba1\u7b97\u4f59\u5f26\u76f8 \u4f3c\u5ea6\uff0c\u8fdb\u884c\u4e3b\u4ece\u4e8b\u4ef6\u5212\u5206\u53ca\u878d\u5408\u5171\u6307\u5143\u7d20\uff0c\u5f97\u5230\u7bc7\u7ae0\u7ea7\u7ed3\u6784\u5316\u4e8b\u4ef6\u4fe1\u606f\u3002\u5b9e\u9a8c\u8bc1\u660eDLEMC\u7684 \u6027\u80fd\u4e0e\u57fa\u51c6\u6a21\u578b\u76f8\u6bd4\u6709\u660e\u663e\u7684\u63d0\u5347\u3002 \u672c\u6587\u7684\u8d21\u732e\u603b\u7ed3\u5982\u4e0b\uff1a (1)\u672c\u6587\u4f9d\u636e\u91d1\u878d\u9886\u57df\u4e0a\u5e02\u516c\u53f8\u516c\u544a\u7ec4\u7ec7\u4e86\u4e00\u4e2a\u7bc7\u7ae0\u7ea7\u522b\u4e8b\u4ef6\u62bd\u53d6\u8bed\u6599\u96c6\uff0c\u5e76\u9488\u5bf9\u8be5\u8bed\u6599\u5b9a \u4e49\u4e86\u4e8b\u4ef6\u53ca\u4e8b\u4ef6\u8868\u793a\u6846\u67b6\u3002 (2)\u672c\u6587\u6784\u5efa\u4e86\u7bc7\u7ae0\u7ea7\u522b\u7684\u4e8b\u4ef6\u62bd\u53d6\u6a21\u578b\uff0c\u8be5\u6a21\u578b\u91c7\u7528\u7aef\u5230\u7aef\u65b9\u5f0f\u8fdb\u884c\u8054\u5408\u5b66\u4e60\uff0c\u540c\u65f6\u5bf9\u4e8b \u4ef6\u68c0\u6d4b\u3001\u4e8b\u4ef6\u5143\u7d20\u8bc6\u522b\u53ca\u5143\u7d20\u89d2\u8272\u5206\u7c7b\u8fdb\u884c\u8bad\u7ec3\uff0c\u901a\u8fc7\u5b9e\u9a8c\u8bc1\u660e\u4e86\u8be5\u6a21\u578b\u7684\u6709\u6548\u6027\u3002 (3)\u672c\u6587\u5728\u4e8b\u4ef6\u5143\u7d20\u8bc6\u522b\u4e2d\u5f15\u5165\u4e8b\u4ef6\u7c7b\u578b\u7279\u5f81\uff0c\u4ee5\u63d0\u9ad8\u4e0d\u540c\u7c7b\u578b\u4e8b\u4ef6\u4e0b\u7684\u5143\u7d20\u8bc6\u522b\u80fd\u529b\u3002\u4e3a \u4e86\u66f4\u51c6\u786e\u8bc6\u522b\u591a\u4e8b\u4ef6\u6587\u6863\u4e2d\u5143\u7d20\u89d2\u8272\uff0c\u6211\u4eec\u5c06\u4e8b\u4ef6\u68c0\u6d4b\u5c42\u8f93\u51fa\u7684\u4e8b\u4ef6\u7c7b\u578b\u53ca\u4e8b\u4ef6\u5143\u7d20\u7684\u5d4c\u5165\u8868 \u793a\u4f5c\u4e3a\u6ce8\u610f\u529b\u5f15\u5165\u6a21\u578b\u7684\u89d2\u8272\u5206\u7c7b\u5c42\uff0c\u4ee5\u63d0\u9ad8\u4e0d\u540c\u4e8b\u4ef6\u7c7b\u578b\u4e0b\u5143\u7d20\u89d2\u8272\u8bc6\u522b\u7684\u51c6\u786e\u7387\u3002 2 \u4e8b \u4e8b \u4e8b\u4ef6 \u4ef6 \u4ef6\u5b9a \u5b9a \u5b9a\u4e49 \u4e49 \u4e49\u53ca \u53ca \u53ca\u8868 \u8868 \u8868\u793a \u793a \u793a \u968f\u7740\u91d1\u878d\u79d1\u6280\u7684\u53d1\u5c55\uff0c\u5728\u91d1\u878d\u9886\u57df\u6bcf\u5929\u90fd\u6709\u6d77\u91cf\u7684\u6570\u636e\u4ea7\u751f\uff0c\u91d1\u878d\u4e8b\u4ef6\u62bd\u53d6\u7814\u7a76\u80fd\u591f\u5e2e\u52a9 \u4eba\u4eec\u8fdb\u884c\u91d1\u878d\u98ce\u9669\u76d1\u63a7\u3001\u8f85\u52a9\u6295\u8d44\u51b3\u7b56\u3001\u5927\u6570\u636e\u5206\u6790\u7b49\u3002\u672c\u6587\u7814\u7a76\u7684\u4e8b\u4ef6\u62bd\u53d6\u9700\u8981\u63d0\u53d6\u4e8b\u4ef6\u7c7b \u522b\u53ca\u53c2\u4e0e\u8005\uff0c\u76ee\u524d\u8be5\u9886\u57df\u7684\u4e2d\u6587\u4e8b\u4ef6\u62bd\u53d6\u7814\u7a76\u7f3a\u4e4f\u6570\u636e\u652f\u6301\uff0c\u524d\u4eba\u7684\u76f8\u5173\u7814\u7a76\u5927\u591a\u6ca1\u6709\u516c\u5f00\u6570 \u636e\u96c6\uff0c\u7814\u7a76\u7684\u4e8b\u4ef6\u7c7b\u578b\u6bd4\u8f83\u96c6\u4e2d\u4e14\u7c7b\u522b\u8f83\u5c11(Yang (2018)4\u7c7b\uff0cZheng (2019)5\u7c7b\uff0c\u53bb\u9664\u91cd\u590d\u540e \u51715\u7c7b\u3002)\uff0c\u4e3a\u6269\u5145\u91d1\u878d\u9886\u57df\u4e8b\u4ef6\u7814\u7a76\u7684\u6570\u636e\uff0c\u672c\u6587\u7ec4\u7ec7\u6784\u5efa\u4e86\u4e00\u5b9a\u89c4\u6a21\u7684\u91d1\u878d\u9886\u57df\u4e2d\u6587\u7bc7\u7ae0\u4e8b \u4ef6\u62bd\u53d6\u6570\u636e\u96c6\uff0c\u5e76\u4f9d\u636e\u81ea\u52a8\u5185\u5bb9\u62bd\u53d6(Automatic Context Extraction\uff0cACE) 0 \u5b9a\u4e49\u7684\u4e8b\u4ef6\u62bd\u53d6 \u4efb\u52a1\uff0c\u8bf4\u660e\u5982\u4e0b\uff1a \u4e8b \u4e8b \u4e8b\u4ef6 \u4ef6 \u4ef6(event)\uff1a\u5728\u67d0\u4e2a\u65f6\u95f4\u70b9\u6216\u65f6\u95f4\u6bb5\uff0c\u4e00\u4e2a\u6216\u591a\u4e2a\u673a\u6784\u7684\u91d1\u878d\u4ea7\u54c1\u7684\u72b6\u6001\u4e3b\u52a8\u5730\u6216\u88ab\u52a8\u5730 \u53d1\u751f\u4e86\u53d8\u5316\u3002 \u5b9e \u5b9e \u5b9e\u4f53 \u4f53 \u4f53(entity)\uff1a\u8bed\u4e49\u7c7b\u522b\u4e2d\u7684\u4e00\u7c7b\u6216\u4e00\u7ec4\u5bf9\u8c61\uff0c\u672c\u6587\u8ba8\u8bba\u7684\u5b9e\u4f53\u5305\u62ec\u547d\u540d\u5b9e\u4f53\u3001\u91d1\u878d\u4ea7\u54c1\u3001 \u65f6\u95f4\u548c\u6570\u503c\u3002 \u4e8b \u4e8b \u4e8b\u4ef6 \u4ef6 \u4ef6\u5143 \u5143 \u5143\u7d20 \u7d20 \u7d20(event argument)\uff1a\u5728\u4e8b\u4ef6\u4e2d\u5177\u6709\u7279\u5b9a\u4f5c\u7528\u7684\u5b9e\u4f53\u3002 \u5143 \u5143 \u5143\u7d20 \u7d20 \u7d20\u89d2 \u89d2 \u89d2\u8272 \u8272 \u8272(argument role)\uff1a\u4e8b\u4ef6\u5143\u7d20\u5728\u4e8b\u4ef6\u4e2d\u627f\u62c5\u7684\u89d2\u8272\u3002 \u9488\u5bf9\u672c\u6587\u7814\u7a76\u7684\u91d1\u878d\u516c\u544a\u4fe1\u606f\uff0c\u4e8b\u4ef6\u5b9a\u4e49\u4e3aevent = def (T, O, F, D, N ) \u5176\u4e2dT \u4e3a\u4e8b\u4ef6\u7c7b \u578b\uff0cO\u3001F \u3001D\u3001N \u4e3a\u4e8b\u4ef6\u4e2d\u76844\u7c7b\u89d2\u8272\uff0c\u5206\u522b\u8868\u793a\u7ec4\u7ec7\u673a\u6784\u3001\u91d1\u878d\u4ea7\u54c1\u3001\u65f6\u95f4\u3001\u6570\u503c\uff0c\u6bcf\u4e00\u7c7b \u4e0b\u6709\u82e5\u5e72\u5c0f\u7c7b\uff0c\u5171\u8ba122\u7c7b\u4e8b\u4ef6\u89d2\u8272\u3002", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "\u56fe1", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "(1)\u4e8b\u4ef6\u7c7b\u578b\uff0c\u4e8b\u4ef6\u6240\u5c5e\u7684\u7c7b\u522b\uff0c\u5982\"\u4e34\u65f6\u505c\u724c\"\u3001\"\u590d\u724c\"\u3001\"\u4e0a\u5e02\u4ea4\u6613\"\u7b49\u3002 (2)\u7ec4\u7ec7\u673a\u6784\uff0c\u53c2\u4e0e\u4e8b\u4ef6\u7684\u4e00\u7c7b\u5b9e\u4f53\uff0c\u5982\"\u4e1c\u6ca3\u79d1\u6280\u96c6\u56e2\u80a1\u4efd\u6709\u9650\u516c\u53f8\"\u3002 (3)\u91d1\u878d\u4ea7\u54c1\uff0c\u91d1\u878d\u9886\u57df\u4e2d\u7684\u76f8\u5173\u4ea7\u54c1\uff0c\u5982\"\u8bc1\u5238\u540d\u79f0\"\uff0c\"\u8bc1\u5238\u7b80\u79f0\"", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "\u56fe1", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "EQUATION", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [ | |
| { | |
| "start": 0, | |
| "end": 8, | |
| "text": "EQUATION", | |
| "ref_id": "EQREF", | |
| "raw_str": "\u3002 (4)\u65f6 \u95f4 \uff0c \u6307 \u4e8b \u4ef6 \u53d1 \u751f \u7684 \u5177 \u4f53 \u65f6 \u95f4 \u70b9 \u6216 \u8005 \u4e8b \u4ef6 \u6301 \u7eed \u53d1 \u751f \u4e14 \u4ea7 \u751f \u4f5c \u7528 \u7684 \u65f6 \u95f4 \u95f4 \u9694 \uff0c \u5982\"10\u65f600\u520601\u79d2\"\u3002 (5)\u6570 \u503c \uff0c \u8861 \u91cf \u4e8b \u4ef6 \u4e2d \u67d0 \u4e00 \u5c5e \u6027 \u5177 \u4f53 \u91cf \u7684 \u591a \u5c11 \uff0c \u5982\"\u7968 \u9762 \u5229 \u73874%\"\u3001\"\u6807 \u51c6 \u4ea4 \u6613 \u5355 \u4f4d10\u5f20\"\u7b49\u3002 \u5176\u4e2d\u4e8b\u4ef6\u7c7b\u578b\u3001\u7ec4\u7ec7\u673a\u6784\u3001\u91d1\u878d\u4ea7\u54c1\u548c\u65f6\u95f4\u7684\u5b9e\u4f8b\u5728\u4e8b\u4ef6\u6587\u672c\u4e2d\u4e00\u5b9a\u4f1a\u51fa\u73b0\uff0c\u6570\u503c\u5b9e\u4f8b\u4e0d \u4e00\u5b9a\u4f1a\u51fa\u73b0\u3002 3 \u4e8b \u4e8b \u4e8b\u4ef6 \u4ef6 \u4ef6\u62bd \u62bd \u62bd\u53d6 \u53d6 \u53d6\u6a21 \u6a21 \u6a21\u578b \u578b \u578bDLEMC \u7bc7\u7ae0\u7ea7\u4e8b\u4ef6\u62bd\u53d6\u7814\u7a76\u8bc6\u522b\u6587\u6863\u4e2d\u5b58\u5728\u7684\u4e8b\u4ef6\u548c\u76f8\u5173\u5143\u7d20\uff0c\u5e76\u5224\u65ad\u5143\u7d20\u5728\u4e8b\u4ef6\u4e2d\u626e\u6f14\u7684 \u89d2\u8272\u3002\u7ed9\u5b9a\u6587\u6863\u96c6doc = {s 0 , s 1 , ..., s Ns }\uff0c\u6bcf\u7bc7\u6587\u6863\u5305\u542b\u6807\u9898\u53e5s 0 \u548c\u5185\u5bb9\u53e5{s 1 , ..., s Ns }\uff0cN s \u4e3a \u53e5\u5b50\u6570\u91cf\u3002\u6a21\u578bDLEMC\u9996\u5148\u5bf9\u6587\u6863\u6807\u9898\u548c\u5185\u5bb9\u5206\u522b\u8fdb\u884c\u4e8b\u4ef6\u68c0\u6d4b\uff0c\u5f97\u5230\u6587\u6863\u5305\u542b\u7684\u4e8b\u4ef6\u7c7b \u578b{t 0 , t 1 , t 2 , . . . },\u5176\u4e2dt 0 \u4e3a\u6807\u9898\u4e2d\u7684\u4e8b\u4ef6\uff0c\u5176\u4ed6\u4e3a\u5185\u5bb9\u4e2d\u7684\u4e8b\u4ef6\uff0c\u7136\u540e\u8bc6\u522b\u51fa\u6587\u6863\u5185\u5bb9\u4e2d\u6bcf\u7c7b\u4e8b \u4ef6\u7684\u76f8\u5173\u5143\u7d20{e 1 , e 2 , . . . }\u53ca\u5176\u5bf9\u5e94\u7684\u89d2\u8272{role 1 , role 2 , . . . }\u3002 DLEMC\u6a21\u578b\u75314\u90e8\u5206\u7ec4\u6210(\u5982\u56fe2\u6240\u793a)\uff0c\u5305\u62ec\u8f93\u5165\u7f16\u7801\u5c42\u3001\u4e8b\u4ef6\u68c0\u6d4b\u5c42\u3001\u4e8b\u4ef6\u5143\u7d20\u8bc6\u522b \u5c42\u548c\u5143\u7d20\u89d2\u8272\u8bc6\u522b\u5c42\u3002 \u8f93 \u8f93 \u8f93\u5165 \u5165 \u5165\u7f16 \u7f16 \u7f16\u7801 \u7801 \u7801\u5c42 \u5c42 \u5c42\uff0c\u57fa\u4e8eBERT\u5bf9\u8f93\u5165\u7684\u53e5\u5b50\u8fdb\u884c\u7f16\u7801\uff0c\u5f97\u5230\u53e5\u5b50\u5bf9\u5e94\u7684\u5411\u91cf\u4ee5\u53ca\u53e5\u5b50\u4e2d\u6bcf \u4e2atoken\u7684\u5411\u91cf\u3002 0 http://projects.ldc.upenn.edu/ace BERT \u4e8b\u4ef6\u5143\u7d20\u8bc6\u522b\u5c42 O B I I O B I Dense+Softmax \u5b9e\u4f53\u5d4c\u5165\u63d0\u53d6 e 1 e 2 \u5b9e\u4f53 q 1 q 2 s 1 \"N\u534e\u76db\u660c\"(002980)\u76d8\u4e2d\u6210\u4ea4\u4ef7... s 2 \u672c\u6240\u81ea\u4eca\u65e509\u65f630\u520600\u79d2\u8d77\u5bf9\u8be5\u80a1... ... ... \u67e5\u8be2\u5411\u91cf \u5143\u7d20\u89d2\u8272\u8bc6\u522b\u5c42 \u4e8b\u4ef6\u68c0\u6d4b\u5c42 \u5173\u4e8e\"N\u534e\u76db\u660c\"\u76d8\u4e2d\u4e34\u65f6\u505c\u724c\u7684\u516c\u544a s 0 token\u5d4c\u5165 \u4e8b\u4ef6\u7c7b\u578b\u5d4c\u5165 Role \u56fe2.\u7bc7\u7ae0\u7ea7\u4e8b\u4ef6\u62bd\u53d6\u6a21\u578bDLEMC \u4e8b \u4e8b \u4e8b\u4ef6 \u4ef6 \u4ef6\u68c0 \u68c0 \u68c0\u6d4b \u6d4b \u6d4b\u5c42 \u5c42 \u5c42\uff0c\u5c06\u7f16\u7801\u5c42\u8f93\u51fa\u7684\u53e5\u5411\u91cf\u4f5c\u4e3a\u8f93\u5165\uff0c\u9884\u6d4b\u8be5\u53e5\u4e2d\u5305\u542b\u7684\u4e8b\u4ef6\uff0c\u4e00\u4e2a\u53e5\u5b50\u4e2d\u53ef\u80fd \u5b58\u5728\u591a\u4e2a\u4e8b\u4ef6\u3002 \u4e8b \u4e8b \u4e8b\u4ef6 \u4ef6 \u4ef6\u5143 \u5143 \u5143\u7d20 \u7d20 \u7d20\u8bc6 \u8bc6 \u8bc6\u522b \u522b \u522b\u5c42 \u5c42 \u5c42\uff0c\u8bc6\u522b\u53e5\u5b50\u4e2d\u53c2\u4e0e\u4e8b\u4ef6\u7684\u5b9e\u4f53\u3002\u5c06\u53e5\u5b50\u7684token\u5411\u91cf\u4e0e\u4e8b\u4ef6\u7c7b\u578b\u5bf9\u5e94\u7684\u5411 \u91cf\u8fdb\u884c\u62fc\u63a5\u4f5c\u4e3a\u8f93\u5165\uff0c\u9884\u6d4b\u6bcf\u4e2atoken\u5bf9\u5e94\u7684BIO\u6807\u7b7e\uff0c\u4ece\u800c\u8bc6\u522b\u51fa\u4e8b\u4ef6\u5143\u7d20\u5bf9\u5e94\u7684\u5b9e\u4f53\u3002 \u5143 \u5143 \u5143\u7d20 \u7d20 \u7d20\u89d2 \u89d2 \u89d2\u8272 \u8272 \u8272\u8bc6 \u8bc6 \u8bc6\u522b \u522b \u522b\u5c42 \u5c42 \u5c42\uff0c\u5bf9\u4e0a\u4e00\u6b65\u8bc6\u522b\u51fa\u7684\u786e\u5b9a\u4e8b\u4ef6\u7c7b\u578b\u4e0b\u7684\u5b9e\u4f53\u8fdb\u884c\u89d2\u8272\u5206\u7c7b\u3002\u5c06\u4e8b\u4ef6\u7c7b\u578bt\u548c \u5b9e\u4f53e\u5bf9\u5e94\u7684\u5d4c\u5165\u8868\u793a\u6c42\u5e73\u5747\u4e4b\u540e\u4f5c\u4e3a\u6ce8\u610f\u529b\u7684\u67e5\u8be2\u5411\u91cf\uff0c\u91cd\u65b0\u8ba1\u7b97token\u7684\u5411\u91cf\u8868\u793a\uff0c\u518d\u5bf9\u6bcf \u4e2a\u4e8b\u4ef6\u5143\u7d20\u7684\u89d2\u8272\u8fdb\u884c\u8bc6\u522b\u3002 \u6a21\u578b\u8bad\u7ec3\u65f6\u5206\u522b\u8ba1\u7b97\u4e8b\u4ef6\u68c0\u6d4b\u5c42\u3001\u4e8b\u4ef6\u5143\u7d20\u8bc6\u522b\u5c42\u4ee5\u53ca\u5143\u7d20\u89d2\u8272\u8bc6\u522b\u5c42\u7684\u635f\u5931\uff0c\u5e76\u5c06\u4e09\u8005 \u6c42\u548c\u4f5c\u4e3a\u6a21\u578b\u6700\u7ec8\u7684\u4f18\u5316\u76ee\u6807\u3002 3.1 \u8f93 \u8f93 \u8f93\u5165 \u5165 \u5165\u7f16 \u7f16 \u7f16\u7801 \u7801 \u7801\u5c42 \u5c42 \u5c42 \u672c\u6587\u91c7\u7528\u9884\u8bad\u7ec3\u8bed\u8a00\u6a21\u578bBERT\u5bf9\u6587\u6863\u8fdb\u884c\u7f16\u7801\uff0c\u8003\u8651BERT\u6a21\u578b\u7684\u6709\u6548\u4f4d\u7f6e\u7f16\u7801\u5e8f\u5217\u957f\u5ea6 \u4ee5\u53ca\u5b9e\u9645\u8bad\u7ec3\u7684\u6a21\u578b\u89c4\u6a21\uff0c\u6211\u4eec\u8bbe\u7f6e\u6700\u5927\u5e8f\u5217\u957f\u5ea6\u4e3amax length\u3002\u5bf9\u4e8e\u7ed9\u5b9a\u7684\u6587\u6863\uff0c\u5c06\u6807\u9898\u4f5c \u4e3a\u72ec\u7acb\u7684\u53e5\u5b50\uff1b\u5bf9\u4e8e\u6587\u6863\u5185\u5bb9\uff0c\u82e5\u6587\u672c\u5e8f\u5217\u957f\u5ea6\u5927\u4e8emax length\uff0c\u5219\u4f9d\u636e\u4e2d\u6587\u6807\u70b9\u7b26\u53f7\u5c06\u5176\u5207 \u5206\u6210\u591a\u4e2a\u53e5\u5b50\uff0c\u53cd\u4e4b\u5219\u5c06\u6574\u7bc7\u6587\u6863\u4f5c\u4e3a\u4e00\u4e2a\u53e5\u5b50\u3002 \u7ecf\u8fc7\u4ee5\u4e0a\u5904\u7406\uff0c\u672c\u6587\u5c06\u4e00\u7bc7\u6587\u6863\u8868\u793a\u4e3a\u4e00\u7cfb\u5217\u53e5\u5b50\u96c6\u5408doc = {s 0 , s 1 , . . . , s Ns }\uff0cN s \u4e3a \u53e5 \u5b50 \u603b \u6570 \uff0cs j \u4e3a \u6587 \u6863 \u4e2d \u7b2cj\u4e2a \u53e5 \u5b50 \uff0cs 0 \u4e3a \u6587 \u6863 \u6807 \u9898 \u3002 \u6bcf \u4e2a \u53e5 \u5b50 \u7531 \u4e00 \u7cfb \u5217token\u7ec4 \u6210{tok 1,j , . . . , tok Nw,j }\uff0c\u5176\u4e2dtok i,j \u4e3a\u7b2cj\u4e2a\u53e5\u5b50\u4e2d\u7b2ci\u4e2atoken\uff0cN w \u4e3a\u7b2cj\u4e2a\u53e5\u5b50\u7684\u5e8f\u5217\u957f\u5ea6\u3002\u6bcf \u4e2a\u53e5\u5b50\u7ecfBERT\u7f16\u7801\u540e\u5f97\u5230\u7684token\u5411\u91cf\u5e8f\u5217\u4e3aH tok = {h 1,j , h 2,j , . . . , h Nw,j }\uff0c\u5176\u4e2dh i,j \u4e3a\u7b2cj\u4e2a \u53e5\u5b50\u4e2d\u7b2ci\u4e2atoken\u5bf9\u5e94\u7684\u5411\u91cf\uff0c\u7ef4\u5ea6\u4e3ad\uff0c\u53e5\u5b50\u5411\u91cf\u5e8f\u5217\u4e3a{h 0 , h 1 , . . . , h Ns }\uff0ch 0 \u4e3a\u6807\u9898\u53e5\u5411\u91cf \u8868\u793a\uff0ch j \u4e3a\u6587\u6863\u4e2d\u7b2cj\u4e2a\u53e5\u5b50\u7684\u5411\u91cf\u8868\u793a\uff0c\u7ef4\u5ea6\u4e3ad\u3002 3.2 \u4e8b \u4e8b \u4e8b\u4ef6 \u4ef6 \u4ef6\u68c0 \u68c0 \u68c0\u6d4b \u6d4b \u6d4b\u5c42 \u5c42 \u5c42 \u4e8b\u4ef6\u68c0\u6d4b\u7684\u76ee\u7684\u662f\u68c0\u6d4b\u53e5\u5b50\u4e2d\u5305\u542b\u7684\u4e8b\u4ef6\uff0c\u672c\u6587\u7684\u6570\u636e\u96c6\u6837\u672c\u4e2d\u53ef\u80fd\u5b58\u5728\u591a\u4e2a\u4e8b\u4ef6\uff0c \u53d7 (Liu al et., 2019)\u542f\u53d1\uff0c\u6211\u4eec\u5c06\u4e8b\u4ef6\u68c0\u6d4b\u5efa\u6a21\u4e3a\u591a\u6807\u7b7e\u5206\u7c7b\u4efb\u52a1\u3002\u4e8b\u4ef6\u68c0\u6d4b\u6837\u672c\u7684\u6807\u6ce8\u5f62\u5f0f \u5982\u88681\u6240\u793a\u3002 \u5176\u4e2ds\u4e3a\u53e5\u5b50\uff0ct 1 \u3001t 2 \u4e3a\u4e0d\u540c\u7684\u4e8b\u4ef6\u7c7b\u578b\uff0c\u6807\u7b7e\u4e3a1\u8868\u793a\u53e5\u5b50s\u4e2d\u5305\u542b\u4e8b\u4ef6t 1 \uff0c\u6807\u7b7e\u4e3a0\u5219\u8868\u793a \u53e5\u5b50s\u4e0d\u5305\u542b\u5bf9\u5e94\u7684\u4e8b\u4ef6\u3002 \u53e5 \u53e5 \u53e5\u5b50 \u5b50 \u5b50 \u4e8b \u4e8b \u4e8b\u4ef6 \u4ef6 \u4ef6\u7c7b \u7c7b \u7c7b\u578b \u578b \u578b \u6807 \u6807 \u6807\u7b7e \u7b7e \u7b7e s t 1 1 s t 2 0 s t 3 1 \u88681.\u4e8b\u4ef6\u68c0\u6d4b\u6570\u636e\u6807\u6ce8\u5b9e\u4f8b \u5bf9\u4e8e\u7ed9\u5b9a\u6587\u6863\u7684\u53e5\u5b50\u5411\u91cf\u8868\u793a{h 0 , h 1 , ..., h Ns }\uff0c\u6211\u4eec\u4f9d\u6b21\u5c06\u6587\u6863\u4e2d\u7684\u53e5\u5b50\u5411\u91cf\u4f5c\u4e3a\u5168\u8fde\u63a5 \u5c42\u7684\u8f93\u5165\uff0c\u5982\u5f0f(1)\u6240\u793a\u3002 H ed = W ed h j + b ed (1) \u5176\u4e2dW ed \u4e3a\u53c2\u6570\u77e9\u9635\uff0cb ed \u4e3a\u504f\u7f6e\uff0ch j \u4e3a\u7b2cj\u4e2a\u53e5\u5b50s j \u7684\u9690\u5c42\u5411\u91cf\u8868\u793a\u3002\u5bf9\u6240\u6709\u4e8b\u4ef6\u7c7b\u578b\u4f7f \u7528Sigmoid\u5206\u7c7b\u5668\u8fdb\u884c\u5206\u7c7b\uff0c\u5f0f(2)\u7ed9\u51fa\u4e86\u5bf9\u67d0\u7c7b\u4e8b\u4ef6\u9884\u6d4b\u7684\u8ba1\u7b97\u65b9\u6cd5\u3002 y = 1 1 + e \u2212H ed (2) \u6b64\u5c42\u7684\u9884\u6d4b\u9519\u8bef\u4f7f\u7528\u4ea4\u53c9\u71b5\u4f5c\u4e3a\u635f\u5931\u51fd\u6570\uff0c\u5982\u5f0f(3)\u6240\u793a\u3002 L ed = \u2212 1 N N j=1 M m=1 y j,m log y * j,m + (1 \u2212 y j,m ) log(1 \u2212 y * j,m ) (3) \u5176\u4e2dy * j,m \u4e3a\u7b2cj\u4e2a\u6837\u672c\u9884\u6d4b\u4e3a\u7b2cm\u7c7b\u4e8b\u4ef6\u7684\u9884\u6d4b\u503c\uff0cy j,m \u4e3a\u7b2cj\u4e2a\u6837\u672c\u4e3a\u7b2cm\u7c7b\u4e8b\u4ef6\u7684\u771f\u5b9e \u503c\uff0cN \u4e3a\u6837\u672c\u603b\u6570\uff0cM \u4e3a\u9884\u5b9a\u4e49\u4e8b\u4ef6\u7c7b\u522b\u603b\u6570\uff0c\u8fd9\u91cc\u8bbe\u5b9a\u4e00\u4e2a\u9608\u503c\u4e3a\u03b1\uff0c\u82e5y * j,m \u7684\u503c\u5927\u4e8e\u7b49 \u4e8e\u03b1\uff0c\u5219\u8ba4\u4e3a\u8be5\u6837\u672c\u5305\u542by * j,m \u5bf9\u5e94\u7684\u4e8b\u4ef6\uff0c\u5426\u5219\u8ba4\u4e3a\u8be5\u6837\u672c\u4e0d\u5305\u542by * j,m \u5bf9\u5e94\u7684\u4e8b\u4ef6\u3002 3.3 \u4e8b \u4e8b \u4e8b\u4ef6 \u4ef6 \u4ef6\u5143 \u5143 \u5143\u7d20 \u7d20 \u7d20\u8bc6 \u8bc6 \u8bc6\u522b \u522b \u522b\u5c42 \u5c42 \u5c42 \u672c \u6587 \u5c06 \u4e8b \u4ef6 \u5143 \u7d20 \u8bc6 \u522b \u5efa \u6a21 \u4e3a \u5e8f \u5217 \u6807 \u6ce8 \u4efb \u52a1 \uff0c \u4f7f \u7528BIO\u6807 \u7b7e \u6a21 \u5f0f (Begin\uff1a \u5b57 \u6bb5 \u5f00 \u5934\uff0cInside\uff1a\u5b57\u6bb5\u5185\u90e8\uff0cOutside\uff1a\u5176\u4ed6\u5b57\u6bb5)\u4e3a\u6bcf\u4e2atoken\u8d4b\u4e88\u4e00\u4e2a\u5b9e\u4f53\u6807\u7b7e\u3002 \u5bf9\u4e8e\u7ed9\u5b9a\u6587\u6863\u4e2d\u7684\u53e5\u5b50s j \uff0c\u9996\u5148\u901a\u8fc7\u4e8b\u4ef6\u68c0\u6d4b\u5c42\u9884\u6d4b\u5f97\u5230\u5bf9\u5e94\u7684\u6587\u6863\u6807\u9898\u4e8b\u4ef6\u548c\u6587\u6863\u5185\u5bb9 \u4e8b\u4ef6\uff0c\u7136\u540e\u4f9d\u6b21\u8bc6\u522b\u6587\u6863\u5185\u5bb9\u4e2d\u6bcf\u4e2a\u4e8b\u4ef6\u7684\u76f8\u5173\u5143\u7d20\u3002\u4e3a\u63d0\u9ad8\u4e0d\u540c\u4e8b\u4ef6\u7c7b\u578b\u4e0b\u5b9e\u4f53\u7684\u8bed\u4e49\u8868 \u793a\uff0c\u672c\u6587\u5728\u5b9e\u4f53\u8bc6\u522b\u90e8\u5206\u5f15\u5165\u4e8b\u4ef6\u7279\u5f81\u3002\u5177\u4f53\u505a\u6cd5\u5982\u4e0b\uff1a\u4e3a\u6bcf\u79cd\u4e8b\u4ef6\u7c7b\u578b\u5b9a\u4e49d\u7ef4(\u4e0etoken\u5411 \u91cf\u7ef4\u5ea6\u76f8\u540c)\u7684\u5411\u91cf\uff0c\u901a\u8fc7\u67e5\u8868\u7684\u65b9\u5f0f\u5f97\u5230\u4e8b\u4ef6\u7c7b\u578b\u5bf9\u5e94\u7684\u5411\u91cft vec \uff0c\u5c06\u53e5\u5b50\u4e2d\u7684\u6bcf\u4e2atoken\u5411 \u91cf\u4e0e\u4e8b\u4ef6\u7c7b\u578b\u5411\u91cft vec \u8fdb\u884c\u62fc\u63a5\u4f5c\u4e3a\u6700\u7ec8\u7684\u7279\u5f81\u5411\u91cf\uff0c\u5f0f(4)\u6240\u793a\u4e3a\u8ba1\u7b97\u67d0token\u7684\u7279\u5f81\u5411\u91cf\u3002 h vec,i = [h i,j ; t vec ]", | |
| "eq_num": "(4)" | |
| } | |
| ], | |
| "section": "\u56fe1", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "\u5176\u4e2d\uff0ch i,j \u4e3a\u7b2cj\u4e2a\u53e5\u5b50\u4e2d\u7b2ci\u4e2atoken\u7684\u5411\u91cf\u8868\u793a\uff0ch vec,i \u4e3a\u53e5\u5b50\u4e2d\u7b2ci\u4e2atoken\u6700\u7ec8\u7684\u7279\u5f81 \u5411\u91cf, \";\"\u8868\u793a\u62fc\u63a5\uff0c\u5c06\u7279\u5f81\u5411\u91cfH vec = {h vec,1 , h vec,2 , . . . , h vec,Nw }\u4f5c\u4e3a\u5168\u8fde\u63a5\u5c42\u7684\u8f93\u5165\uff0c\u4f7f \u7528sof tmax\u4f5c\u4e3a\u5206\u7c7b\u5668\u9884\u6d4b\u6bcf\u4e2atoken\u5bf9\u5e94\u7684\u6807\u7b7e\uff0c\u5982\u5f0f(5)\u6240\u793a\u3002 P = sof tmax(W ner H vec + b ner )", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "\u56fe1", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "W ner \u4e3a\u53c2\u6570\u77e9\u9635\uff0cb ner \u4e3a\u504f\u7f6e\u3002\u4f7f\u7528\u4ea4\u53c9\u71b5\u8ba1\u7b97\u8be5\u90e8\u5206\u7684\u635f\u5931\uff0c\u5982\u5f0f(6)\u6240\u793a\u3002", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "\u56fe1", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "EQUATION", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [ | |
| { | |
| "start": 0, | |
| "end": 8, | |
| "text": "EQUATION", | |
| "ref_id": "EQREF", | |
| "raw_str": "L ner = \u2212 1 N N i=1 K k=1 y i,k log P i,k", | |
| "eq_num": "(6)" | |
| } | |
| ], | |
| "section": "\u56fe1", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "EQUATION", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [ | |
| { | |
| "start": 0, | |
| "end": 8, | |
| "text": "EQUATION", | |
| "ref_id": "EQREF", | |
| "raw_str": "N \u4e3a\u6837\u672c\u603b\u6570\uff0cK\u4e3a\u6807\u7b7e\u7c7b\u522b\u603b\u6570\uff0c\u7b2ci\u4e2a\u6837\u672c\u9884\u6d4b\u4e3a\u7b2cK\u4e2a\u6807\u7b7e\u7684\u6982\u7387\u4e3aP i,k \uff0c\u7b2ci\u4e2a\u6837\u672c \u771f\u5b9e\u7684\u6807\u7b7e\u4e3ay i,k \u3002 3.4 \u5143 \u5143 \u5143\u7d20 \u7d20 \u7d20\u89d2 \u89d2 \u89d2\u8272 \u8272 \u8272\u5206 \u5206 \u5206\u7c7b \u7c7b \u7c7b\u5c42 \u5c42 \u5c42 \u5143\u7d20\u89d2\u8272\u8bc6\u522b\u7684\u76ee\u6807\u662f\u4e3a\u786e\u5b9a\u4e8b\u4ef6\u7c7b\u578b\u4e0b\u7684\u5b9e\u4f53\u8d4b\u4e88\u9884\u5b9a\u4e49\u7684\u89d2\u8272\uff0c\u672c\u6587\u5c06\u89d2\u8272\u8bc6\u522b\u5efa\u6a21 \u4e3a\u591a\u5206\u7c7b\u4efb\u52a1\u3002\u4e3a\u66f4\u597d\u5730\u533a\u5206\u5b9e\u4f53\u626e\u6f14\u7684\u89d2\u8272\uff0c\u5c06\u5229\u7528\u6ce8\u610f\u529b\u673a\u5236\u6765\u589e\u5f3a\u6587\u672c\u7684\u7279\u5f81\u8868\u793a\uff0c\u4f9d \u6b21\u5224\u65ad\u6bcf\u4e2a\u5b9e\u4f53\u5728\u4e8b\u4ef6\u4e2d\u626e\u6f14\u7684\u89d2\u8272\u3002 \u5b9e\u4f53\u5f80\u5f80\u5305\u542b\u591a\u4e2atoken\uff0c\u5bf9\u4e8e\u7ed9\u5b9a\u53e5\u5b50s j \u4e2d\u8bc6\u522b\u51fa\u7684\u5b9e\u4f53\u96c6E = {e 1 , e 2 , . . . }\uff0c\u5176\u4e2d\u6bcf\u4e2a \u5b9e\u4f53\u5305\u542b\u8be5\u53e5\u4e2d\u7684\u7b2ci\u81f3\u7b2ck\u4e2atoken,[tok i,j , . . . , tok k,j ]\uff0c\u672c\u6587\u53d6\u5b9e\u4f53\u4e2d\u7684\u6240\u6709\u5b57\u7b26\u5411\u91cf\u7684\u5747\u503c\u4f5c \u4e3a\u8be5\u5b9e\u4f53\u7684\u5d4c\u5165\u8868\u793ac\uff0c\u7ef4\u5ea6\u4e3ad\uff0c\u91c7\u7528\u8fd9\u79cd\u5747\u503c\u5411\u91cf\u53ef\u4ee5\u6709\u6548\u907f\u514d\u6a21\u578b\u8fc7\u62df\u5408\u95ee\u9898 (Ji al et., 2018)\u3002\u8ba1\u7b97\u65b9\u5f0f\u5982\u5f0f(7)\u6240\u793a\uff0ch i,j \u4e3a\u7b2cj\u4e2a\u53e5\u5b50\u4e2d\u7684\u7b2ci\u4e2atoken\u3002 c = 1 k k i=1 h i,j", | |
| "eq_num": "(7)" | |
| } | |
| ], | |
| "section": "\u56fe1", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "\u7136\u540e\uff0c\u6211\u4eec\u5c06E vec \u4e2d\u7684\u7b2cj\u4e2a\u5b9e\u4f53\u7684\u5411\u91cf\u8868\u793ac j \u4e0e\u5305\u542b\u8be5\u5b9e\u4f53\u7684\u4e8b\u4ef6\u7c7b\u578b\u5bf9\u5e94\u7684\u5411\u91cf\u8868 \u793at vec \u8fdb\u884c\u76f8\u52a0\u518d\u6c42\u5e73\u5747\u5f97\u5230\u7ef4\u5ea6\u4e3ad\u7684,\u6ce8\u610f\u529b\u673a\u5236\u67e5\u8be2\u5411\u91cfq\uff0c\u5982\u5f0f(8)\u6240\u793a\u3002", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "\u56fe1", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "EQUATION", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [ | |
| { | |
| "start": 0, | |
| "end": 8, | |
| "text": "EQUATION", | |
| "ref_id": "EQREF", | |
| "raw_str": "q = t vec + c j 2 (8) \u6700\u7ec8\u5f97\u5230\u67e5\u8be2\u5411\u91cf\u96c6\u5408Q = q 1 , q 2 , . . . \uff0c\u4f7f\u7528q\u4e0e\u53e5\u5b50s j \u7684token\u5411\u91cf\u8868\u793aH tok \u8ba1\u7b97\u5f97\u5230\u6bcf \u4e2atoken\u7684\u6ce8\u610f\u529b\u503ca k \uff0c\u8ba1\u7b97\u65b9\u6cd5\u5982\u5f0f(9)\u6240\u793a\u3002 a k = exp(h k,j q T l,j ) Ns i=1 exp(h i,j q T l,j )", | |
| "eq_num": "(9)" | |
| } | |
| ], | |
| "section": "\u56fe1", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "\u5176 \u4e2d \uff0cq T l,j \u4e3a \u53e5 \u5b50s j \u7b2cl\u67e5 \u8be2 \u5411 \u91cf \uff0ch k,j \u8868 \u793a \u53e5 \u5b50s j \u4e2d \u7b2ck\u4e2atoken\u7684 \u5bf9 \u5e94 \u7684 \uff0ca k \u4e3a \u7b2ck\u4e2atoken\u7684\u6ce8\u610f\u529b\u503c\uff0ca = [a 1 , a 2 , . . . , a Ns ]\u4e3a\u6240\u6709token\u7684\u6ce8\u610f\u529b\u503c\u5411\u91cf\uff0c\u7531a\u4e0e\u6837\u672c\u4e2d\u6bcf \u4e2atoken\u7684\u5411\u91cfH tok \u76f8\u4e58\u5f97\u5230\u878d\u5408\u5b9e\u4f53\u4fe1\u606f\u548c\u4e8b\u4ef6\u7c7b\u578b\u8bed\u4e49\u4fe1\u606f\u7684\u7279\u5f81\u5411\u91cfV \uff0c\u5982\u516c\u5f0f(10)\u6240 \u793a\uff0c * \u8868\u793a\u70b9\u4e58\u3002 V = a * H tok (10)", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "\u56fe1", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "\u5c06V \u4f5c\u4e3a\u5168\u8fde\u63a5\u5c42\u7684\u8f93\u5165\uff0c\u4f7f\u7528sof tmax\u5206\u7c7b\u5668\u8fdb\u884c\u5206\u7c7b\uff0c\u5982\u516c\u5f0f(11)\u6240\u793a\u3002 ", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "\u56fe1", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "EQUATION", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [ | |
| { | |
| "start": 0, | |
| "end": 8, | |
| "text": "EQUATION", | |
| "ref_id": "EQREF", | |
| "raw_str": "y * = sof tmax(W rt V + b rt ) (11) \u5176\u4e2dM \u4e3a\u7c7b\u522b\u603b\u6570\uff0cy m j,u \u4e3a\u7b2cj\u4e2a\u6837\u672c\u4e2d\u7b2cu\u4e2a\u4e8b\u4ef6\u5143\u7d20\u5c5e\u4e8e\u7b2cj\u7c7b\u9884\u5b9a\u4e49\u89d2\u8272\u7c7b\u578b\u7684\u771f\u5b9e \u503c\uff0cy * (m) j,u \u4e3a\u6a21\u578b\u9884\u6d4b\u503c, N \u4e3a\u6837\u672c\u603b\u6570\u3002\u6b64\u90e8\u5206\u7684\u635f\u5931\u51fd\u6570\u5982(12)\u6240\u793a\uff1a L rt = \u2212 1 N N j=1 u=1 M m=1 y m j,u log y * (m) j,u", | |
| "eq_num": "(12" | |
| } | |
| ], | |
| "section": "\u56fe1", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "EQUATION", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [ | |
| { | |
| "start": 0, | |
| "end": 8, | |
| "text": "EQUATION", | |
| "ref_id": "EQREF", | |
| "raw_str": "score list = sim(t 1,i , t 2,j ) i, j \u2208 {1, 2, ...} t 1,i , t 2,j \u2208 (T, O, F, D, N )", | |
| "eq_num": "(14)" | |
| } | |
| ], | |
| "section": "\u56fe1", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "\u5f0f(14)\u4e2dt 1,i \u4e0et 2,j \u5206\u522b\u8868\u793a\u4e8b\u4ef6e 1 \u7684\u7b2ci\u4e2a\u5143\u7d20\u4e0e\u4e8b\u4ef6e 2 \u7684\u7b2cj\u4e2a\u5143\u7d20\uff0cscore list \u4e3a\u8ba1\u7b97\u5f97\u5230\u7684 \u8bed\u4e49\u76f8\u4f3c\u5ea6\u7684\u5f97\u5206\u96c6\u5408\uff0c\u82e5\u8be5\u96c6\u5408\u4e2d\u67d0\u4e00\u9879\u503c\u5927\u4e8e\u9608\u503c\u03b3\uff0c\u5c31\u5c06\u8be5\u9879\u5bf9\u5e94\u7684\u4e24\u4e2a\u4e8b\u4ef6\u5143\u7d20\u5408\u5e76\u3002 ", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "\u56fe1", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "4 \u5b9e \u5b9e \u5b9e\u9a8c \u9a8c \u9a8c 4.1 \u6570 \u6570 \u6570\u636e \u636e \u636e\u96c6 \u96c6 \u96c6 \u672c\u6587\u5c06\u4ece\u4e92\u8054\u7f51\u4e0a\u641c\u96c6\u7684\u4e0a\u5e02\u516c\u53f8\u516c\u544a\u4f5c\u4e3a\u5b9e\u9a8c\u6570\u636e\u96c6\u3002\u5171\u6709\u6587\u6863\u603b\u657023067\uff0c\u5176\u4e2d5056\u4e2a \u6587 \u6863 \u4e2d \u5305 \u542b \u591a \u4e2a \u4e8b \u4ef6 \uff0c \u5360 \u6bd421.9%\uff0c \u5c06 \u603b \u6587 \u6863 \u6309 \u71678\uff1a1\uff1a1\u5212 \u5206 \u6210 \u8bad \u7ec3 \u96c6 \u3001 \u9a8c \u8bc1 \u96c6 \u548c \u6d4b \u8bd5 \u96c6\uff0c\u6570\u636e\u96c6\u4e2d\u6807\u6ce8\u7684\u5b9e\u4f53\u7c7b\u522b\u5305\u62ecNUM(Number\uff0c\u6570\u503c)\u3001ORG(Organization\uff0c\u7ec4\u7ec7\u673a \u6784)\u3001FIN(Finance\uff0c\u91d1\u878d\u4ea7\u54c1)\u3001TIM(Time\uff0c\u65e5\u671f\u3001\u65f6\u95f4)\u3002\u4e8b\u4ef6\u5206\u4e3a11\u7c7b\uff1a\u4e0a\u5e02\u4ea4\u6613\u3001 \u505c\u724c\u3001\u4e34\u65f6\u505c\u724c\u3001\u590d\u724c\u3001\u6458\u724c\u3001\u540d\u79f0\u53d8\u66f4\u3001\u652f\u4ed8\u5229\u606f\u3001\u503a\u5238\u8f6c\u8ba9\u3001\u6682\u505c\u4e0a\u5e02\u3001\u7ec8\u6b62\u4e0a\u5e02\u3001\u5230\u671f \u5151\u4ed8\u3002\u5404\u7c7b\u6837\u672c\u6570\u5982\u88682\u6240\u793a\u3002 \u4e8b \u4e8b \u4e8b\u4ef6 \u4ef6 \u4ef6\u7c7b \u7c7b \u7c7b\u578b \u578b \u578b \u8bad \u8bad \u8bad\u7ec3 \u7ec3 \u7ec3\u96c6 \u96c6 \u96c6 \u9a8c \u9a8c \u9a8c\u8bc1 \u8bc1 \u8bc1\u96c6 \u96c6 \u96c6 \u6d4b \u6d4b \u6d4b\u8bd5 \u8bd5 \u8bd5\u96c6 \u96c6 \u96c6 \u603b \u603b \u603b\u6570 \u6570 \u6570 \u591a \u591a \u591a\u4e8b \u4e8b \u4e8b\u4ef6 \u4ef6 \u4ef6\u7387 \u7387 \u7387(%", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "\u56fe1", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "EQUATION", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [ | |
| { | |
| "start": 0, | |
| "end": 8, | |
| "text": "EQUATION", | |
| "ref_id": "EQREF", | |
| "raw_str": "F 1 = (2 * P * R) (P +R) 4.3 \u53c2 \u53c2 \u53c2\u6570 \u6570 \u6570\u8bbe \u8bbe \u8bbe\u7f6e \u7f6e \u7f6e \u672c\u6587\u7684\u5b9e\u9a8c\u57fa\u4e8eBERT base\u6a21\u578b\u6765\u521d\u59cb\u5316\u8bcd\u5411\u91cf\uff0c\u7ef4\u5ea6\u4e3a786\uff0cdropout\u7684\u6bd4\u7387\u4e3a0.4\uff0c\u6279\u6b21 \u5927\u5c0f\u4e3a16\uff0c\u6a21\u578b\u5b66\u4e60\u7387\u4e3a3e-5\uff0c\u8bad\u7ec310\u4e2aepoch\uff0c\u6700\u5927\u6587\u672c\u5e8f\u5217\u957f\u5ea6max length\u4e3a200\uff0c\u4e8b\u4ef6\u7c7b \u578b\u5411\u91cf\u968f\u673a\u521d\u59cb\u5316\u751f\u6210\uff0c\u7ef4\u5ea6\u4e3a768\u3002 4.4 \u7ed3 \u7ed3 \u7ed3\u679c \u679c \u679c\u5206 \u5206 \u5206\u6790 \u6790 \u6790 \u6211 \u4eec \u5b9e \u73b0 \u4e86Yang (2018)\u63d0 \u51fa \u7684DCFEE\u6a21 \u578b \uff0cZheng (2019)\u63d0 \u51fa \u7684Doc2EDAG\u6a21 \u578b \uff0c \u4ee5 \u53ca\u57fa\u4e8eBERT\u7684\u7ba1\u9053\u6a21\u578bBERT-P\uff0c\u5728BERT-P\u4e2d\u4e8b\u4ef6\u68c0\u6d4b\u90e8\u5206\u4e0e\u672c\u6587\u63d0\u51fa\u7684\u65b9\u6cd5DLEMC\u76f8 \u540c \uff0c \u4f46 \u5728 \u5b9e \u4f53 \u8bc6 \u522b \u4efb \u52a1 \u4e2d \u672a \u589e \u52a0 \u4e8b \u4ef6 \u7279 \u5f81 \uff0c \u5728 \u89d2 \u8272 \u5206 \u7c7b \u4efb \u52a1 \u4e2d \u672a \u5229 \u7528 \u4e8b \u4ef6 \u548c \u5b9e \u4f53 \u7279 \u5f81 \u6ce8\u610f\u529b\uff0cDLEMC-P\u4e3aDLEMC\u7684\u7ba1\u9053\u6a21\u5f0f\u3002\u6211\u4eec\u5c06DCFEE\u3001BERT-P\u548cDoc2EDAG\u4f5c\u4e3a\u672c\u6587 \u7684baseline\uff0c\u5728\u5305\u542b\u5168\u90e811\u7c7b\u4e8b\u4ef6\u7684\u6d4b\u8bd5\u96c6\u4e0a\u8fdb\u884c\u5404\u9879\u6d4b\u8bd5\u3002 1)\u4e3a\u9a8c\u8bc1\u672c\u6587\u7684\u6a21\u578bDLEMC\u5728\u4e8b\u4ef6\u7c7b\u578b\u68c0\u6d4b\u4e0a\u7684\u6709\u6548\u6027\uff0c\u6211\u4eec\u5728\u6d4b\u8bd5\u96c6\u4e0a\u5bf9\u6a21\u578b\u8fdb\u884c\u4e86 \u8bc4\u4ef7\uff0c\u5b9e\u9a8c\u7ed3\u679c\u5982\u88684\u6240\u793a\u3002 Model P(%) R(%) F1(%)", | |
| "eq_num": "DCFEE 80" | |
| } | |
| ], | |
| "section": "\u56fe1", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "\u5982\u88684\u6240\u793a\uff0c\u5229\u7528\u5e8f\u5217\u6807\u6ce8\u8bc6\u522b\u4e8b\u4ef6\u89e6\u53d1\u8bcd\u7684DCFEE\u6a21\u578b\u6548\u679c\u6bd4\u8f83\u5dee\u3002\u901a\u8fc7\u5206\u7c7b\u6a21\u578b \u68c0\u6d4b\u4e8b\u4ef6\u7684BERT-P\u3001DLEMC-P\u3001DLEMC\u3001Doc2EDAG\u6a21\u578b\u5728\u5404\u9879\u6307\u6807\u4e0a\u5747\u4f18\u4e8eDCFEE\uff0c \u5176 \u4e2d \u7aef \u5230 \u7aef \u8054 \u5408 \u5b66 \u4e60 \u6a21 \u578bDLEMC\u3001Doc2EDAG\u5728 \u5404 \u9879 \u6307 \u6807 \u4e0a \u5747 \u4f18 \u4e8e \u7ba1 \u9053 \u5f0f \u6a21 \u578b \u3002 \u672c \u6587 \u63d0 \u51fa\u7684DLEMC\u6a21\u578b\u5728\u51c6\u786e\u7387\u4e0a\u7565\u4f4e\u4e8e\u91c7\u7528\u4e8c\u5206\u7c7b\u7b56\u7565\u7684Doc2EDAG\u6a21\u578b\uff0c\u4f46\u53ec\u56de\u7387\u548cF1\u5747\u4f18 \u4e8eDoc2EDAG\uff0c\u5176\u4e2dF1\u63d0\u9ad8\u4e860.2%\uff0c\u5b9e\u9a8c\u8868\u660e\u591a\u6807\u7b7e\u5206\u7c7b\u6a21\u578b\u5728\u591a\u4e8b\u4ef6\u68c0\u6d4b\u4e2d\u6709\u8f83\u597d\u7684\u8868\u73b0\u3002", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "\u56fe1", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "2)\u4e3a\u9a8c\u8bc1DLEMC\u6a21\u578b\u5728\u7bc7\u7ae0\u4e8b\u4ef6\u5143\u7d20\u8bc6\u522b\u548c\u89d2\u8272\u5206\u7c7b\u65f6\u7684\u6709\u6548\u6027\uff0c\u6211\u4eec\u5728\u6d4b\u8bd5\u96c6\u4e0a\u5bf9\u6a21 \u578b\u8fdb\u884c\u4e86\u8bc4\u4ef7\u3002 \u7531\u88685\u5b9e\u9a8c\u7ed3\u679c\u53ef\u4ee5\u770b\u51fa\uff0c\u57fa\u4e8e\u9884\u8bad\u7ec3\u8bed\u8a00\u6a21\u578bBERT\u8bcd\u5411\u91cf\u8868\u5f81\u7684\u4e8b\u4ef6\u62bd\u53d6\u6a21\u578bBERT-P\u660e\u663e\u4f18\u4e8e\u4f7f\u7528word2vec (Mikolov et al., 2013 ", | |
| "cite_spans": [ | |
| { | |
| "start": 108, | |
| "end": 129, | |
| "text": "(Mikolov et al., 2013", | |
| "ref_id": "BIBREF14" | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "\u56fe1", | |
| "sec_num": null | |
| } | |
| ], | |
| "back_matter": [ | |
| { | |
| "text": "\u5728\u7279\u5b9a\u9886\u57df\u6709\u8f83\u9ad8\u7684\u51c6\u786e\u7387\uff0c\u4f46\u662f\u901a\u5e38\u9700\u8981\u7f16\u5199\u5927\u91cf\u7684\u4eba\u5de5\u6a21\u677f\uff0c\u4e14\u666e\u9002\u6027\u8f83\u5dee (Yangarber and Grishman, 1997; Surdeanu and Harabagiu, 2002) \u3002\u7edf\u8ba1\u5b66\u4e60\u65b9\u6cd5\u53ef\u4ee5\u5206\u4e3a\u4e24\u7c7b:\u4f20\u7edf\u7684\u57fa\u4e8e \u7279\u5f81\u5de5\u7a0b\u7684\u673a\u5668\u5b66\u4e60\u65b9\u6cd5\u4ee5\u53ca\u57fa\u4e8e\u6df1\u5ea6\u5b66\u4e60\u7684\u65b9\u6cd5\u3002\u4f20\u7edf\u7279\u5f81\u5de5\u7a0b\u4e3b\u8981\u4f9d\u8d56\u81ea\u7136\u8bed\u8a00\u5904\u7406\u5de5\u5177 \u83b7\u53d6\u6709\u6548\u7684\u7279\u5f81(\u5982\u53e5\u6cd5\u3001\u8bcd\u6c47\u3001\u8bcd\u6027\u7b49)\uff0c\u7136\u540e\u5229\u7528\u4f20\u7edf\u7684\u5206\u7c7b\u6a21\u578b(\u5982\u6700\u5927\u71b5\u3001\u652f\u6301\u5411\u91cf \u673a)\u8fdb\u884c\u5206\u7c7b (Ahn, 2006; Jungermann and Morik, 2008; Liao and Grishman, 2010) \u3002\u57fa\u4e8e\u6df1\u5ea6 \u5b66\u4e60\u7684\u65b9\u6cd5\u4f9d\u9760\u795e\u7ecf\u7f51\u7edc\u81ea\u52a8\u63d0\u53d6\u7279\u5f81\uff0c\u5728\u4e8b\u4ef6\u62bd\u53d6\u4e2d\u53d6\u5f97\u4e86\u4e0d\u9519\u7684\u6548\u679c\u3002\u5982WU (2019)\u4f7f\u7528 \u4e00\u79cd\u6df7\u5408\u795e\u7ecf\u7f51\u7edc\u6a21\u578b\uff0c\u8fdb\u884c\u5b9e\u4f53\u548c\u4e8b\u4ef6\u7684\u8054\u5408\u5b66\u4e60\u3002Chen (2015)\u4f7f\u7528\u4e00\u79cd\u52a8\u6001\u591a\u6c60\u5316\u5377\u79ef\u795e \u7ecf\u7f51\u7edc\u6765\u6355\u83b7\u591a\u4e2a\u7279\u5f81\uff0c\u63d0\u5347\u4e86\u4e8b\u4ef6\u62bd\u53d6\u7684\u6027\u80fd\u3002Zeng (2016)\u4f7f\u7528\u4e00\u79cd\u5377\u79ef\u53cc\u5411LSTM\u795e\u7ecf\u7f51 \u7edc\uff0c\u5206\u522b\u4ece\u8bcd\u7ea7\u522b\u548c\u5b57\u7ea7\u522b\u8fdb\u884c\u89e6\u53d1\u8bcd\u548c\u5b9e\u4f53\u7684\u8bc6\u522b\u3002 \u4e8b\u4ef6\u62bd\u53d6\u4efb\u52a1\u4f9d\u636e\u662f\u5426\u5177\u6709\u9884\u5b9a\u4e49\u7684\u4e8b\u4ef6\u6846\u67b6(\u4e8b\u4ef6\u7c7b\u578b\u53ca\u6bcf\u7c7b\u4e8b\u4ef6\u5bf9\u5e94\u7684\u89d2\u8272)\u53ef\u4ee5\u5206 \u4e3a\u5f00\u653e\u57df\u548c\u9650\u5b9a\u57df\u4e8b\u4ef6\u62bd\u53d6 (Wei and Wang, 2019)\uff0c\u5f00\u653e\u57df\u4e8b\u4ef6\u62bd\u53d6\u76ee\u6807\u5728\u4e8e\u8bc6\u522b\u81ea\u7136\u8bed\u8a00\u6587 \u672c\u4e2d\u7684\u4e8b\u4ef6\uff0c\u4e00\u822c\u6ca1\u6709\u9886\u57df\u9650\u5236\uff0c\u4e0d\u9700\u8981\u9884\u5b9a\u4e49\u4e8b\u4ef6\u6846\u67b6\u3002\u9650\u5b9a\u57df\u4e8b\u4ef6\u62bd\u53d6\u4f1a\u9884\u5148\u5b9a\u4e49\u597d\u8981\u62bd \u53d6\u7684\u4e8b\u4ef6\u7c7b\u578b\uff0c\u5982\"\u88ad\u51fb\"\u4e8b\u4ef6\u3001\"\u5ba1\u5224\"\u4e8b\u4ef6\u7b49\u3002\u540c\u65f6\u4e5f\u4f1a\u5b9a\u4e49\u6bcf\u7c7b\u4e8b\u4ef6\u53c2\u4e0e\u8005\u7684\u89d2\u8272\uff0c\u5982\"\u5ba1 \u5224\"\u4e8b\u4ef6\u4e2d\u5305\u542b\"\u88ab\u5ba1\u5224\u4eba\"\u3001\"\u5ba1\u5224\u65f6\u95f4\"\u3001\"\u5730\u70b9\"\u7b49\u89d2\u8272\u3002 ", | |
| "cite_spans": [ | |
| { | |
| "start": 36, | |
| "end": 66, | |
| "text": "(Yangarber and Grishman, 1997;", | |
| "ref_id": "BIBREF15" | |
| }, | |
| { | |
| "start": 67, | |
| "end": 96, | |
| "text": "Surdeanu and Harabagiu, 2002)", | |
| "ref_id": "BIBREF16" | |
| }, | |
| { | |
| "start": 208, | |
| "end": 219, | |
| "text": "(Ahn, 2006;", | |
| "ref_id": null | |
| }, | |
| { | |
| "start": 220, | |
| "end": 247, | |
| "text": "Jungermann and Morik, 2008;", | |
| "ref_id": "BIBREF17" | |
| }, | |
| { | |
| "start": 248, | |
| "end": 272, | |
| "text": "Liao and Grishman, 2010)", | |
| "ref_id": "BIBREF18" | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "\u76ee\u524d\u4e8b\u4ef6\u62bd\u53d6\u65b9\u6cd5\u53ef\u4ee5\u5206\u4e3a\u4e24\u7c7b\uff1a\u57fa\u4e8e\u6a21\u5f0f\u5339\u914d\u65b9\u6cd5\u548c\u57fa\u4e8e\u7edf\u8ba1\u5b66\u4e60\u65b9\u6cd5\u3002\u6a21\u5f0f\u5339\u914d\u65b9\u6cd5", | |
| "sec_num": null | |
| } | |
| ], | |
| "bib_entries": { | |
| "BIBREF0": { | |
| "ref_id": "b0", | |
| "title": "Online news event extraction for global crisis surveillance", | |
| "authors": [ | |
| { | |
| "first": "J", | |
| "middle": [], | |
| "last": "Piskorski", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "H", | |
| "middle": [], | |
| "last": "Tanev", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "M", | |
| "middle": [], | |
| "last": "Atkinson", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "E", | |
| "middle": [ | |
| "V D" | |
| ], | |
| "last": "Goot", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "", | |
| "middle": [], | |
| "last": "Zavarella", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2011, | |
| "venue": "Transactions on computational collective intelligence", | |
| "volume": "6910", | |
| "issue": "1", | |
| "pages": "182--212", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "J.Piskorski,H.Tanev,M.Atkinson,E.V.D.Goot,andV.Zavarella. 2011. Online news event extraction for global crisis surveillance, Transactions on computational collective intelligence, vol. 6910, no. 1, pp. 182-212.", | |
| "links": null | |
| }, | |
| "BIBREF1": { | |
| "ref_id": "b1", | |
| "title": "Unsupervised event clustering and aggregation from news wire and web articles", | |
| "authors": [ | |
| { | |
| "first": "S", | |
| "middle": [], | |
| "last": "Ribeiro", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "O", | |
| "middle": [], | |
| "last": "Ferret", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "", | |
| "middle": [], | |
| "last": "Tannier", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2017, | |
| "venue": "Proceedings of the 2017 EMNLP Workshop: Natural Language Processing meets Journalism", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "62--67", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "S.Ribeiro,O.Ferret,andX.Tannier. 2017. Unsupervised event clustering and aggregation from news wire and web articles, in Proceedings of the 2017 EMNLP Workshop: Natural Language Processing meets Journalism, pp. 62-67.", | |
| "links": null | |
| }, | |
| "BIBREF2": { | |
| "ref_id": "b2", | |
| "title": "Exploiting structured news information to improve event detection via dual-level clustering", | |
| "authors": [ | |
| { | |
| "first": "S", | |
| "middle": [], | |
| "last": "Yu", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "B", | |
| "middle": [], | |
| "last": "Wu", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2018, | |
| "venue": "IEEE Third International Conference on Data Science in Cyberspace", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "873--880", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "S. Yu and B. Wu. 2018. Exploiting structured news information to improve event detection via dual-level clustering, in IEEE Third International Conference on Data Science in Cyberspace, pp. 873-880.", | |
| "links": null | |
| }, | |
| "BIBREF3": { | |
| "ref_id": "b3", | |
| "title": "Event extraction via dynamic multipooling convolutional neural networks", | |
| "authors": [ | |
| { | |
| "first": "Y", | |
| "middle": [], | |
| "last": "Chen", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "L", | |
| "middle": [], | |
| "last": "Xu", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "K", | |
| "middle": [], | |
| "last": "Liu", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "D", | |
| "middle": [], | |
| "last": "Zeng", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "J", | |
| "middle": [], | |
| "last": "Zhao", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2015, | |
| "venue": "Proceedings of the 53rd Annual Meeting of the Association for Computational Linguistics and the 7th International Joint Conference on Natural Language Processing", | |
| "volume": "1", | |
| "issue": "", | |
| "pages": "167--176", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Chen, Y., Xu, L., Liu, K., Zeng, D., Zhao, J. 2015. Event extraction via dynamic multipooling con- volutional neural networks. In: Proceedings of the 53rd Annual Meeting of the Association for Computational Linguistics and the 7th International Joint Conference on Natural Language Pro- cessing, vol. 1, pp. 167-176.", | |
| "links": null | |
| }, | |
| "BIBREF4": { | |
| "ref_id": "b4", | |
| "title": "A convolution BiLSTM neural network model for Chinese event extraction", | |
| "authors": [ | |
| { | |
| "first": "Y", | |
| "middle": [], | |
| "last": "Zeng", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "H", | |
| "middle": [], | |
| "last": "Yang", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Y", | |
| "middle": [], | |
| "last": "Feng", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Z", | |
| "middle": [], | |
| "last": "Wang", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "D", | |
| "middle": [], | |
| "last": "Zhao", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2016, | |
| "venue": "Natural Language Understanding and Intelligent Applications", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "275--287", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Zeng, Y., Yang, H., Feng, Y., Wang, Z., Zhao, D. 2016. A convolution BiLSTM neural network model for Chinese event extraction. In Natural Language Understanding and Intelligent Applications (pp. 275-287). Springer, Cham.", | |
| "links": null | |
| }, | |
| "BIBREF5": { | |
| "ref_id": "b5", | |
| "title": "Joint event extraction via structured prediction with global features", | |
| "authors": [ | |
| { | |
| "first": "Qi", | |
| "middle": [], | |
| "last": "Li", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Ji", | |
| "middle": [], | |
| "last": "Heng", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Liang", | |
| "middle": [], | |
| "last": "Huang", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2013, | |
| "venue": "ACL", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Qi Li, Heng Ji, and Liang Huang. 2013. Joint event extraction via structured prediction with global features. In ACL.", | |
| "links": null | |
| }, | |
| "BIBREF6": { | |
| "ref_id": "b6", | |
| "title": "Joint Event Extraction via Recurrent Neural Networks", | |
| "authors": [ | |
| { | |
| "first": "Thien", | |
| "middle": [], | |
| "last": "Nguyen", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "", | |
| "middle": [], | |
| "last": "Huu", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Kyunghyun", | |
| "middle": [], | |
| "last": "Cho", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Ralph", | |
| "middle": [], | |
| "last": "Grishman", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2016, | |
| "venue": "Proceedings of the 2016 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Nguyen, Thien Huu and Cho, Kyunghyun and Grishman, Ralph. 2016. Joint Event Extraction via Recurrent Neural Networks. Proceedings of the 2016 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies.", | |
| "links": null | |
| }, | |
| "BIBREF7": { | |
| "ref_id": "b7", | |
| "title": "DCFEE: A Document-level Chinese Financial Event Extraction System based on Automatically Labeled Training Data", | |
| "authors": [ | |
| { | |
| "first": "Hang", | |
| "middle": [], | |
| "last": "Yang", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Yubo", | |
| "middle": [], | |
| "last": "Chen", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Kang", | |
| "middle": [], | |
| "last": "Liu", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Yang", | |
| "middle": [], | |
| "last": "Xiao", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Jun", | |
| "middle": [], | |
| "last": "Zhao", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2018, | |
| "venue": "Proceedings of ACL 2018", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Hang Yang,Yubo Chen,Kang Liu,Yang Xiao,and Jun Zhao. 2018. DCFEE: A Document-level Chinese Financial Event Extraction System based on Automatically Labeled Training Data, In Proceedings of ACL 2018, System Demonstrations.", | |
| "links": null | |
| }, | |
| "BIBREF8": { | |
| "ref_id": "b8", | |
| "title": "Doc2EDAG: An End-to-End Document-level Framework for Chinese Financial Event Extraction", | |
| "authors": [ | |
| { | |
| "first": "Zheng", | |
| "middle": [], | |
| "last": "Shun", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Cao", | |
| "middle": [], | |
| "last": "Wei", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Xu", | |
| "middle": [], | |
| "last": "Wei", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Bian", | |
| "middle": [], | |
| "last": "Jiang", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2019, | |
| "venue": "Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Zheng Shun,Cao Wei,Xu Wei,Bian Jiang. 2019. Doc2EDAG: An End-to-End Document-level Framework for Chinese Financial Event Extraction, Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP).", | |
| "links": null | |
| }, | |
| "BIBREF9": { | |
| "ref_id": "b9", | |
| "title": "Document-level Event Extraction Based on Joint Labeling and Global Reasoning", | |
| "authors": [ | |
| { | |
| "first": "Yang", | |
| "middle": [], | |
| "last": "Zhong Weifeng", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "", | |
| "middle": [], | |
| "last": "Hang", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "", | |
| "middle": [], | |
| "last": "Chen Yubo", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "", | |
| "middle": [], | |
| "last": "Kang", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "", | |
| "middle": [], | |
| "last": "Zhao Jun", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2019, | |
| "venue": "", | |
| "volume": "33", | |
| "issue": "", | |
| "pages": "88--95", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "ZHONG Weifeng, YANG Hang, CHEN Yubo, LIU Kang, ZHAO Jun. 2019. Document-level Event Extraction Based on Joint Labeling and Global Reasoning.,33(9): 88-95,106.", | |
| "links": null | |
| }, | |
| "BIBREF10": { | |
| "ref_id": "b10", | |
| "title": "BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding", | |
| "authors": [ | |
| { | |
| "first": "J", | |
| "middle": [], | |
| "last": "Devlin", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "M", | |
| "middle": [], | |
| "last": "Chang", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "K", | |
| "middle": [], | |
| "last": "Lee", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "K", | |
| "middle": [], | |
| "last": "Toutanova", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2019, | |
| "venue": "ArXiv", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Devlin,J., Chang,M., Lee,K.,Toutanova,K. 2019. BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding. ArXiv, abs/1810.04805.", | |
| "links": null | |
| }, | |
| "BIBREF11": { | |
| "ref_id": "b11", | |
| "title": "Event Detection without Triggers", | |
| "authors": [ | |
| { | |
| "first": "Shulin &", | |
| "middle": [], | |
| "last": "Liu", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Yang", | |
| "middle": [ | |
| "&" | |
| ], | |
| "last": "Li", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "", | |
| "middle": [], | |
| "last": "Zhang", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "&", | |
| "middle": [], | |
| "last": "Feng", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "", | |
| "middle": [], | |
| "last": "Yang", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Xinpeng", | |
| "middle": [], | |
| "last": "Zhou", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2019, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": { | |
| "DOI": [ | |
| "735-744.10.18653/v1/N19-1080" | |
| ] | |
| }, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Liu, Shulin & Li, Yang & Zhang, Feng & Yang, Tao & Zhou, Xinpeng. 2019. Event Detection without Triggers.735-744. 10.18653/v1/N19-1080.", | |
| "links": null | |
| }, | |
| "BIBREF12": { | |
| "ref_id": "b12", | |
| "title": "Improving Neural Fine-Grained Entity Typing with Knowledge Attention", | |
| "authors": [ | |
| { | |
| "first": "Ji", | |
| "middle": [], | |
| "last": "Xin", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Yankai", | |
| "middle": [], | |
| "last": "Lin", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Zhiyuan", | |
| "middle": [], | |
| "last": "Liu", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Maosong", | |
| "middle": [], | |
| "last": "Sun", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2018, | |
| "venue": "The 32th AAAI Conference on Artificial Intelligence", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Ji Xin, Yankai Lin, Zhiyuan Liu, Maosong Sun. 2018. Improving Neural Fine-Grained Entity Typing with Knowledge Attention. The 32th AAAI Conference on Artificial Intelligence.", | |
| "links": null | |
| }, | |
| "BIBREF13": { | |
| "ref_id": "b13", | |
| "title": "Adam: A Method for Stochastic Optimization", | |
| "authors": [ | |
| { | |
| "first": "Diederik", | |
| "middle": [ | |
| "P" | |
| ], | |
| "last": "Kingma", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Jimmy", | |
| "middle": [], | |
| "last": "Ba", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2015, | |
| "venue": "ICLR", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Kingma, Diederik P. and Ba, Jimmy. 2015. Adam: A Method for Stochastic Optimization. In ICLR.", | |
| "links": null | |
| }, | |
| "BIBREF14": { | |
| "ref_id": "b14", | |
| "title": "Distributed representations of words and phrases and their compositionality", | |
| "authors": [ | |
| { | |
| "first": "T", | |
| "middle": [], | |
| "last": "Mikolov", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "I", | |
| "middle": [], | |
| "last": "Sutskever", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "K", | |
| "middle": [], | |
| "last": "Chen", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2013, | |
| "venue": "Advances in Neural Information Processing Systems", | |
| "volume": "26", | |
| "issue": "", | |
| "pages": "3111--3119", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Mikolov T,Sutskever I,Chen K,et al. 2013. Distributed representations of words and phrases and their compositionality. Advances in Neural Information Processing Systems, 26, 3111-3119.", | |
| "links": null | |
| }, | |
| "BIBREF15": { | |
| "ref_id": "b15", | |
| "title": "Customization of information extraction systems", | |
| "authors": [ | |
| { | |
| "first": "R", | |
| "middle": [], | |
| "last": "Yangarber", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "R", | |
| "middle": [], | |
| "last": "Grishman", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1997, | |
| "venue": "Proceedings of International Workshop on Lexically Driven Information", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "1--11", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Yangarber R, Grishman R. 1997. Customization of information extraction systems. [C]//Proceedings of International Workshop on Lexically Driven Information Extraction\uff1a1-11.", | |
| "links": null | |
| }, | |
| "BIBREF16": { | |
| "ref_id": "b16", | |
| "title": "Infrastructure for open-domain information extraction", | |
| "authors": [ | |
| { | |
| "first": "M", | |
| "middle": [], | |
| "last": "Surdeanu", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "S", | |
| "middle": [ | |
| "M" | |
| ], | |
| "last": "Harabagiu", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2002, | |
| "venue": "ceedings of the 2nd International Conference on Human Language Technology Research", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "325--330", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Surdeanu M,Harabagiu S M. 2002. Infrastructure for open-domain information extraction. [C]// Pro- ceedings of the 2nd International Conference on Human Language Technology Research. Morgan Kaufmann Publishers Inc\uff1a325-330.", | |
| "links": null | |
| }, | |
| "BIBREF17": { | |
| "ref_id": "b17", | |
| "title": "Enhanced Services for Targeted Information Retrieval by Event Extraction and Data Mining", | |
| "authors": [ | |
| { | |
| "first": "Felix", | |
| "middle": [], | |
| "last": "Jungermann", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Katharina", | |
| "middle": [], | |
| "last": "Morik", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2008, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Felix Jungermann, Katharina Morik. 2008. Enhanced Services for Targeted Information Retrieval by Event Extraction and Data Mining. [M]// Natural Language and Information Systems. Springer Berlin Heidelberg.", | |
| "links": null | |
| }, | |
| "BIBREF18": { | |
| "ref_id": "b18", | |
| "title": "Using Document Level Cross-Event Inference to Improve Event Extraction", | |
| "authors": [ | |
| { | |
| "first": "S", | |
| "middle": [], | |
| "last": "Liao", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "R", | |
| "middle": [], | |
| "last": "Grishman", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2010, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Liao S, Grishman R. 2010. Using Document Level Cross-Event Inference to Improve Event Extraction.", | |
| "links": null | |
| }, | |
| "BIBREF19": { | |
| "ref_id": "b19", | |
| "title": "Proceedings of the 48th Annual Meeting of the Association for Computational Linguistics", | |
| "authors": [ | |
| { | |
| "first": "//", | |
| "middle": [], | |
| "last": "Acl", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2010, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "// ACL 2010, Proceedings of the 48th Annual Meeting of the Association for Computational Linguistics, July 11-16, 2010, Uppsala, Sweden. DBLP.", | |
| "links": null | |
| }, | |
| "BIBREF20": { | |
| "ref_id": "b20", | |
| "title": "Joint Extraction of Entities and Events by a Hybrid Neural Network", | |
| "authors": [ | |
| { | |
| "first": "", | |
| "middle": [], | |
| "last": "Wu Wentao", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "", | |
| "middle": [], | |
| "last": "Li Peifeng", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "", | |
| "middle": [], | |
| "last": "Qiaoming", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2019, | |
| "venue": "", | |
| "volume": "33", | |
| "issue": "", | |
| "pages": "77--83", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "WU Wentao, LI Peifeng, ZHU Qiaoming. 2019. Joint Extraction of Entities and Events by a Hybrid Neural Network. 33(8): 77-83.", | |
| "links": null | |
| }, | |
| "BIBREF21": { | |
| "ref_id": "b21", | |
| "title": "Joint extraction of events and entities within a document context", | |
| "authors": [ | |
| { | |
| "first": "Yang", | |
| "middle": [], | |
| "last": "Bishan", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Mitchell", | |
| "middle": [], | |
| "last": "Tom", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "M", | |
| "middle": [], | |
| "last": "", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2016, | |
| "venue": "Proceedings of NAACL-HLT 2016. San Diego\uff0cCalifornia\uff0cUSA: Association for Computational Linguistics", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "289--299", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Yang Bishan and Mitchell Tom M. 2016. Joint extraction of events and entities within a document context. // Proceedings of NAACL-HLT 2016. San Diego\uff0cCalifornia\uff0cUSA: Association for Com- putational Linguistics: 289-299", | |
| "links": null | |
| } | |
| }, | |
| "ref_entries": { | |
| "TABREF0": { | |
| "text": "1 \u5f15 \u5f15 \u5f15\u8a00 \u8a00 \u8a00 \u00a92020 \u4e2d\u56fd\u8ba1\u7b97\u8bed\u8a00\u5b66\u5927\u4f1a \u6839\u636e\u300aCreative Commons Attribution 4.0 International License\u300b\u8bb8\u53ef\u51fa\u7248 \u57fa\u91d1\u9879\u76ee\uff1a\u56fd\u5bb6\u81ea\u7136\u79d1\u5b66\u57fa\u91d1\u9752\u5e74\u9879\u76ee(61906035)\uff0c\u4e0a\u6d77\u5e02\u9752\u5e74\u79d1\u6280\u82f1\u624d\u626c\u5e06\u8ba1\u5212\u9879\u76ee(19YF1402300) \u8fd1\u5e74\u6765\u4e92\u8054\u7f51\u5feb\u901f\u53d1\u5c55\uff0c\u7f51\u7edc\u5a92\u4f53\u6bcf\u5929\u4ea7\u751f\u5927\u91cf\u7684\u65b0\u95fb\u3001\u516c\u544a\u7b49\u975e\u7ed3\u6784\u5316\u4fe1\u606f\u3002\u4fe1\u606f\u62bd\u53d6 \u6280\u672f\u7814\u7a76\u5982\u4f55\u4ece\u6d77\u91cf\u7684\u4fe1\u606f\u4e2d\u5feb\u901f\u6709\u6548\u5730\u6355\u83b7\u6709\u4ef7\u503c\u7684\u4fe1\u606f\uff0c\u4ee5\u5e2e\u52a9\u4eba\u4eec\u9488\u5bf9\u7279\u5b9a\u4fe1\u606f\u505a\u5206 \u6790\u3001\u51b3\u7b56\u3002\u4e8b\u4ef6\u62bd\u53d6\u662f\u4fe1\u606f\u62bd\u53d6\u7684\u5206\u652f\uff0c\u65e8\u5728\u4ece\u975e\u7ed3\u6784\u5316\u7684\u81ea\u7136\u8bed\u8a00\u6587\u672c\u4e2d\u62bd\u53d6\u51fa\u7528\u6237\u611f\u5174\u8da3 \u7684\u4e8b\u4ef6\u4fe1\u606f\u5e76\u4ee5\u7ed3\u6784\u5316\u7684\u5f62\u5f0f\u5c55\u793a (Ahn, 2006)\u3002\u4e8b\u4ef6\u62bd\u53d6\u5728\u5f88\u591a\u9886\u57df\u6709\u7740\u5e7f\u6cdb\u7684\u5e94\u7528\uff0c\u5982\u6784\u5efa \u4e8b\u4ef6\u77e5\u8bc6\u56fe\u8c31\u3001\u4fe1\u606f\u68c0\u7d22\u3001\u81ea\u52a8\u95ee\u7b54\u4ee5\u53ca\u8f85\u52a9\u5176\u4ed6\u81ea\u7136\u8bed\u8a00\u5904\u7406\u4efb\u52a1\u7b49\u3002 \u4e8b\u4ef6\u62bd\u53d6\u5206\u4e3a\u5f00\u653e\u57df\u548c\u9650\u5b9a\u57df\u4e8b\u4ef6\u62bd\u53d6 (Wei and Wang, 2019)\u3002\u5f00\u653e\u57df\u4e8b\u4ef6\u62bd\u53d6\u7814\u7a76\u901a\u5e38 \u6ca1\u6709\u9886\u57df\u8303\u56f4\u9650\u5236\uff0c\u4e8b\u4ef6\u7c7b\u578b\u53ca\u4e8b\u4ef6\u7684\u6846\u67b6\u7ed3\u6784\u672a\u77e5\uff0c\u4e3b\u8981\u5229\u7528\u65e0\u76d1\u7763\u65b9\u6cd5\u4ece\u6587\u672c\u4e2d\u53d1\u73b0\u4e8b \u4ef6", | |
| "html": null, | |
| "content": "<table/>", | |
| "type_str": "table", | |
| "num": null | |
| }, | |
| "TABREF2": { | |
| "text": "\u6211\u4eec\u5c06\u4e8b\u4ef6\u68c0\u6d4b\u3001\u4e8b\u4ef6\u5143\u7d20\u8bc6\u522b\u4ee5\u53ca\u5143\u7d20\u89d2\u8272\u5206\u7c7b\u540c\u65f6\u8fdb\u884c\u8bad\u7ec3\uff0c\u6a21\u578b\u7684\u8bad\u7ec3\u76ee\u6807\u662f\u7efc \u54083\u90e8\u5206\u7684\u635f\u5931\u8fbe\u5230\u6700\u5c0f\uff0c\u8bad\u7ec3\u65f6\u5206\u522b\u8ba1\u7b97\u4e8b\u4ef6\u68c0\u6d4b\u5c42\u7684\u4e8c\u5206\u7c7b\u4ea4\u53c9\u71b5\u635f\u5931L ed \uff0c\u4ee5\u53ca\u4e8b\u4ef6\u5143\u7d20 \u8bc6\u522b\u5c42\u4e0e\u5143\u7d20\u89d2\u8272\u5206\u7c7b\u5c42\u7684\u591a\u5206\u7c7b\u4ea4\u53c9\u71b5\u635f\u5931L ner \u4e0eL rt \uff0c\u6211\u4eec\u5c063\u4e2a\u635f\u5931\u6c42\u548c\u4f5c\u4e3a\u6a21\u578b\u6700\u7ec8\u7684 \u4f18\u5316\u76ee\u6807\uff1aL f inal = L ed + L ner + L rt\u6a21\u578b\u8bad\u7ec3\u65f6\u91c7\u7528Adam(Kingma al et., 2015)\u4f5c\u4e3a\u4f18\u5316\u5668\uff0c\u901a\u8fc7\u9a8c\u8bc1\u96c6\u9009\u62e9\u6700\u597d\u7684\u6a21\u578b\u8fdb\u884c \u9884\u6d4b\u3002 3.6 \u4e3b \u4e3b \u4e3b\u4ece \u4ece \u4ece\u4e8b \u4e8b \u4e8b\u4ef6 \u4ef6 \u4ef6\u8bc6 \u8bc6 \u8bc6\u522b \u522b \u522b \u6587\u6863\u4e2d\u5305\u542b\u591a\u4e2a\u4e8b\u4ef6\u65f6\uff0c\u5c06\u6839\u636e\u6807\u9898\u4e8b\u4ef6\u8fdb\u884c\u4e3b\u4ece\u4e8b\u4ef6\u5212\u5206\uff0c\u7136\u540e\u5bf9\u540c\u6307\u4e8b\u4ef6\u5143\u7d20\u8fdb\u884c\u878d \u5408\uff0c\u4ece\u800c\u5f97\u5230\u7bc7\u7ae0\u7ea7\u4e8b\u4ef6\u62bd\u53d6\u7ed3\u679c\u3002\u6587\u6863\u6807\u9898\u5f80\u5f80\u80fd\u6982\u62ec\u4e00\u7bc7\u6587\u6863\u7684\u4e3b\u8981\u5185\u5bb9\uff0c\u6545\u672c\u6587\u5c06\u6587\u6863 \u6807\u9898\u4e2d\u7684\u4e8b\u4ef6\u4f5c\u4e3a\u4e3b\u4e8b\u4ef6\uff0c\u5176\u4ed6\u4e8b\u4ef6\u4f5c\u4e3a\u4ece\u4e8b\u4ef6\u3002 \u5bf9\u4e8e\u6587\u6863\u4e2d\u7684\u591a\u4e2a\u4e8b\u4ef6Events = {e 0 , e 1 , e 2 , . . . }\uff0c\u5176\u4e2de 0 \u4e3a\u6587\u6863\u6807\u9898\u9884\u6d4b\u51fa\u7684\u4e8b\u4ef6\uff0c\u5176\u4ed6", | |
| "html": null, | |
| "content": "<table><tr><td colspan=\"2\">\u7684\u5747\u4e3a\u6587\u6863\u5185\u5bb9\u9884\u6d4b\u51fa\u7684\u4e8b\u4ef6\u3002\u4e8b\u4ef6\u7c7b\u578b\u4ee5\u53ca\u4e8b\u4ef6\u5143\u7d20\u4f7f\u7528\u4eceDLEMC\u6a21\u578b\u4e2d\u83b7\u5f97\u7684\u5d4c\u5165\u5f0f\u8868</td></tr><tr><td>\u793a\uff0c\u57fa\u4e8e\u4f59\u5f26\u76f8\u4f3c\u5ea6\u8ba1\u7b97\u4e24\u4e2a\u4e8b\u4ef6\u7684\u76f8\u4f3c\u7a0b\u5ea6\uff0c\u5982\u5f0f(13)\u6240\u793a\u3002</td><td/></tr><tr><td>obj = sim(e 0 , e i )</td><td>(13)</td></tr><tr><td colspan=\"2\">obj\u4e3a\u76f8\u4f3c\u5ea6\u5f97\u5206\uff0c\u7528\u4e8e\u8861\u91cfe 0 \u4e0ee i \u4e24\u4e2a\u4e8b\u4ef6\u7684\u5171\u6307\u7a0b\u5ea6\uff0c\u6211\u4eec\u53d6\u6700\u9ad8\u5f97\u5206\u5bf9\u5e94\u7684\u90a3\u7ec4\u4e8b\u4ef6</td></tr><tr><td>\u4f5c\u4e3a\u6587\u6863\u7684\u4e3b\u4e8b\u4ef6\uff0c\u5176\u4ed6\u4e8b\u4ef6\u4e3a\u4ece\u4e8b\u4ef6\u3002</td><td/></tr><tr><td colspan=\"2\">\u4e3b\u4ece\u4e8b\u4ef6\u878d\u5408\u7684\u76ee\u7684\u662f\u5bf9\u540c\u4e00\u4e2a\u6587\u6863\u91cc\u591a\u4e2a\u4e8b\u4ef6\u4e4b\u95f4\u7684\u5171\u6307\u4e8b\u4ef6\u5143\u7d20\u8fdb\u884c\u5408\u5e76\uff0c\u4ece\u800c\u5f97\u5230</td></tr><tr><td colspan=\"2\">\u89c4\u8303\u7684\u7bc7\u7ae0\u7ea7\u4e8b\u4ef6\u4fe1\u606f\u3002\u672c\u6587\u901a\u8fc7\u8ba1\u7b97\u4e0d\u540c\u4e8b\u4ef6\u4e2d\u4e8b\u4ef6\u5143\u7d20\u7684\u8bed\u4e49\u76f8\u4f3c\u5ea6\u6765\u8861\u91cf\u5b83\u4eec\u7684\u5171\u6307\u7a0b</td></tr><tr><td colspan=\"2\">\u5ea6\uff0c\u5177\u4f53\u89c4\u5219\u4e3a\uff1a\u8bed\u4e49\u76f8\u4f3c\u5ea6\u8d85\u8fc7\u8bbe\u5b9a\u9608\u503c\u03b3\u7684\u4e8b\u4ef6\u5143\u7d20\u4f5c\u4e3a\u5171\u6307\u5143\u7d20\uff0c\u5426\u5219\u4e3a\u975e\u5171\u6307\u5143\u7d20\u3002\u4e24</td></tr><tr><td>\u4e2a\u4e8b\u4ef6\u4e2d\u4e0d\u540c\u5143\u7d20\u7684\u76f8\u4f3c\u5ea6\u57fa\u4e8e\u4f59\u5f26\u76f8\u4f3c\u5ea6\u8ba1\u7b97\uff0c\u5982\u516c\u5f0f(14)\u6240\u793a\u3002</td><td/></tr><tr><td/><td>)</td></tr><tr><td>3.5 \u6a21 \u6a21 \u6a21\u578b \u578b \u578b\u8bad \u8bad \u8bad\u7ec3 \u7ec3 \u7ec3</td><td/></tr></table>", | |
| "type_str": "table", | |
| "num": null | |
| } | |
| } | |
| } | |
| } |