| { | |
| "paper_id": "2020", | |
| "header": { | |
| "generated_with": "S2ORC 1.0.0", | |
| "date_generated": "2023-01-19T12:52:50.037476Z" | |
| }, | |
| "title": "Korean Sentence Ordering Based on Sub Word Level Word Vector and Pointer Network", | |
| "authors": [ | |
| { | |
| "first": "Xiaodong", | |
| "middle": [], | |
| "last": "Yan", | |
| "suffix": "", | |
| "affiliation": { | |
| "laboratory": "", | |
| "institution": "Minzu University of China National language resource monitoring &Research Center Minority Languages Branch", | |
| "location": {} | |
| }, | |
| "email": "" | |
| }, | |
| { | |
| "first": "Xiaoqing", | |
| "middle": [], | |
| "last": "Xie", | |
| "suffix": "", | |
| "affiliation": { | |
| "laboratory": "", | |
| "institution": "Minzu University of China", | |
| "location": {} | |
| }, | |
| "email": "" | |
| } | |
| ], | |
| "year": "", | |
| "venue": null, | |
| "identifiers": {}, | |
| "abstract": "Sentence sorting is one of the most important tasks in multi document summarization system and machine reading comprehension. The quality of sorting will directly affect the coherence and readability of abstracts and answers. Therefore, this paper adopts the deep learning method which is widely used in both Chinese and English, combined with the characteristics of the rich morphological changes of Korean words, puts forward a Korean sentence ordering model based on the sub word level word vector and pointer network, the purpose of which is to solve the problem that traditional methods can not mine deep semantic information. In this paper, a morpheme split based word vector training method (morv) is proposed, and the Korean word vector is obtained by \u8ba1\u7b97\u8bed\u8a00\u5b66 comparing the sub word n-ary word vector training method (SG). Two sentence vector methods are used: convolution neural network (CNN) and long-term memory network (LSTM), combined with pointer network. The results show that the combination of morv and LSTM can better capture the semantic logic relationship between sentences and improve the effect of sentence ordering.", | |
| "pdf_parse": { | |
| "paper_id": "2020", | |
| "_pdf_hash": "", | |
| "abstract": [ | |
| { | |
| "text": "Sentence sorting is one of the most important tasks in multi document summarization system and machine reading comprehension. The quality of sorting will directly affect the coherence and readability of abstracts and answers. Therefore, this paper adopts the deep learning method which is widely used in both Chinese and English, combined with the characteristics of the rich morphological changes of Korean words, puts forward a Korean sentence ordering model based on the sub word level word vector and pointer network, the purpose of which is to solve the problem that traditional methods can not mine deep semantic information. In this paper, a morpheme split based word vector training method (morv) is proposed, and the Korean word vector is obtained by \u8ba1\u7b97\u8bed\u8a00\u5b66 comparing the sub word n-ary word vector training method (SG). Two sentence vector methods are used: convolution neural network (CNN) and long-term memory network (LSTM), combined with pointer network. The results show that the combination of morv and LSTM can better capture the semantic logic relationship between sentences and improve the effect of sentence ordering.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Abstract", | |
| "sec_num": null | |
| } | |
| ], | |
| "body_text": [ | |
| { | |
| "text": "EQUATION", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [ | |
| { | |
| "start": 0, | |
| "end": 8, | |
| "text": "EQUATION", | |
| "ref_id": "EQREF", | |
| "raw_str": "3)\u4f7f\u7528\u5f62\u6001\u7d20\u548c\u5b50\u8bcd\u7ea7\u522bn\u5143\u8fdb\u884c\u8bcd\u5411\u91cf\u8bad\u7ec3\uff0c\u5e76\u5bf9\u6bd4\u6548\u679c\uff1b 4)\u4f7f\u7528\u4e24\u79cd\u8bcd\u5411\u91cf\u8bad\u7ec3\u65b9\u6cd5\u5f97\u5230\u8bcd\u5411\u91cf\uff0c\u518d\u4f7f\u7528\u4e24\u79cd\u4e0d\u540c\u7684\u53e5\u5411\u91cf\u8bad\u7ec3\u65b9\u6cd5\u5f97\u5230\u53e5\u5411\u91cf\uff0c \u6700\u540e\u8fdb\u884c\u53e5\u5b50\u6392\u5e8f\u5b9e\u9a8c\uff0c\u5e76\u5bf9\u6bd4\u6548\u679c\u3002 2 \u671d\u9c9c\u8bed\u53e5\u5b50\u6392\u5e8f\u6a21\u578b 2.1 \u4efb\u52a1\u63cf\u8ff0 \u5728\u673a\u5668\u9605\u8bfb\u7406\u89e3\u7684\u7b54\u6848\u878d\u5408\u4efb\u52a1\u548c\u591a\u6587\u6863\u81ea\u52a8\u6587\u6458\u4efb\u52a1\u4e2d\uff0c\u5019\u9009\u7684\u53e5\u5b50\u96c6\u662f\u4ece\u4e0d\u540c\u7684\u6587\u6863 \u4e2d\u62bd\u53d6\u7684\uff0c\u56e0\u6b64\uff0c\u65e0\u6cd5\u6839\u636e\u53e5\u5b50\u5728\u539f\u6587\u4e2d\u7684\u4f4d\u7f6e\u6216\u8005\u4e00\u4e9b\u663e\u5f0f\u7684\u8fde\u63a5\u8bcd\u5bf9\u4e71\u5e8f\u7684\u53e5\u5b50\u96c6\u5408\u8fdb\u884c \u6392\u5e8f\u3002 \u53e5\u5b50\u6392\u5e8f\u4efb\u52a1\u8981\u89e3\u51b3\u7684\u95ee\u9898\u5c31\u662f\u628a\u4e00\u7ec4\u4e71\u5e8f\u7684\u53e5\u5b50\uff0c\u6392\u5217\u6210\u8fde\u8d2f\u3001 \u901a\u987a\u7684\u6bb5\u843d\u3002 \u8bbe\u7ed9\u5b9a \u4e00\u7ec4\u4e71\u5e8f\u7684\u53e5\u5b50S = s 1 , s 2 , \u2022 \u2022 \u2022 , s n \uff0c\u53e5\u5b50\u6392\u5e8f\u7684\u4efb\u52a1\u76ee\u6807\u662f\u5c06\u5176\u6392\u5217\u6210\u987a\u5e8fo * \uff0c\u5bf9\u4e8e\u987a\u5e8fo * \u6709\uff1a s o 1 * > s o 2 * > \u2022 \u2022 \u2022 > s on * (1) \u5728\u7ed9\u5b9a\u53e5\u5b50\u96c6S\u7684\u60c5\u51b5\u4e0b\uff0c \u987a\u5e8fo * \u7684\u6982\u7387P (o * |S)\u5927\u4e8e\u5176\u4ed6\u4efb\u4f55\u987a\u5e8f\u7684\u6982\u7387\uff0c \u53ef\u4ee5\u8868\u793a\u4e3a \u5f0f(2)\u3002\u5176\u4e2do\u8868\u793a\u53e5\u5b50\u96c6S\u7684\u4efb\u4e00\u79cd\u6392\u5e8f\uff0c\u800c\u03a8\u8868\u793a\u53e5\u5b50\u96c6S\u7684\u6240\u6709\u53ef\u80fd\u7684\u6392\u5e8f\u7684\u96c6\u5408\u3002 P (o * |S) > P (o|S), \u2200o \u2208 \u03a8 (2) 2.2 \u6a21\u578b\u67b6\u6784 \u6211\u4eec\u91c7\u7528\u6307\u9488\u7f51\u7edc\u6a21\u578b(Pointer Network)(Vinyals et al., 2015)\u5bf9\u53e5\u5b50\u96c6S\u8fdb\u884c\u6392\u5e8f\u3002\u6307\u9488\u7f51 \u7edc(Pointer Network)\u662fNallapati\u7b49(Nallapati et al., 2016)\u63d0\u51fa\u7684\u57fa\u4e8e\u6ce8\u610f\u529b\u673a\u5236\u7684\u5e8f\u5217\u5230\u5e8f\u5217 \u6a21\u578b\u7684\u4e00\u4e2a\u53d8\u79cd\u3002 \u5b83\u4e0d\u662f\u628a\u4e00\u4e2a\u5e8f\u5217\u8f6c\u6362\u6210\u53e6\u4e00\u4e2a\u5e8f\u5217\uff0c\u800c\u662f\u4ea7\u751f\u4e00\u7cfb\u5217\u6307\u5411\u8f93\u5165\u5e8f\u5217\u5143\u7d20\u7684 \u6307\u9488\u3002\u6700\u57fa\u7840\u7684\u7528\u6cd5\u662f\u5bf9\u53ef\u53d8\u957f\u5ea6\u5e8f\u5217\u6216\u96c6\u5408\u7684\u5143\u7d20\u8fdb\u884c\u6392\u5e8f\uff0c\u4e5f\u9002\u7528\u4e8e\u53e5\u5b50\u6392\u5e8f\u95ee\u9898\u3002 \u56fe 1: \u6307\u9488\u7f51\u7edc\u6a21\u578b\u7ed3\u6784 \u56fe 2: \u57fa\u4e8e\u6307\u9488\u7f51\u7edc\u7684\u53e5\u5b50\u6392\u5e8f\u6a21\u578b \u6307\u9488\u7f51\u7edc\u6a21\u578b\u975e\u5e38\u7b80\u6d01\u5982\u56fe1\u6240\u793a\uff0c\u7ed3\u6784\u662f\u57fa\u672c\u7684seq2seq+attention\u3002\u57fa\u4e8e\u6307\u9488\u7f51\u7edc\u7684\u53e5\u5b50 \u6392\u5e8f\u6a21\u578b\u7684\u67b6\u6784\u5982\u56fe2\u6240\u793a\u3002\u4ee5\u987a\u5e8fo\u4e3a\u96c6\u5408S\u6392\u5e8f\u7684\u6982\u7387P (o|S)\u7684\u8ba1\u7b97\u516c\u5f0f\u4e3a(3)\u3002 P (o|s) = n i=1 P = (o i |o i\u22121 , \u2022 \u2022 \u2022 , o 1 , s)", | |
| "eq_num": "(3)" | |
| } | |
| ], | |
| "section": "", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "EQUATION", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [ | |
| { | |
| "start": 0, | |
| "end": 8, | |
| "text": "EQUATION", | |
| "ref_id": "EQREF", | |
| "raw_str": "\u6982\u7387P = (o i |o i\u22121 , \u2022 \u2022 \u2022 , o 1 , s)\u53ef\u4ee5\u901a\u8fc7\u6307\u9488\u7f51\u7edc\u8ba1\u7b97\uff0c\u4e3a\u5f0f(5)\uff0c(6)\uff0c\u5176\u4e2de j \uff0cd i \u5206\u522b\u662f\u6307\u9488\u7f51\u7edc \u7f16\u7801\u7aef\u548c\u89e3\u7801\u7aef\u7684\u8f93\u51fa\u3002 P = (o i |o i\u22121 , \u2022 \u2022 \u2022 , o 1 , s) = sof tmaxt(u i ) o i (4) u i j = v T tanh(W 1 e j + W 2 d i ) (5) 2.2.1 \u7f16\u7801\u7aef \u6307\u9488\u7f51\u7edc\u7684\u7f16\u7801\u5668\u6a21\u578b\u53ef\u4ee5\u8868\u793a\u4e3a\u5f0f(6)\uff0c\u5176\u4e2d\uff0cEnc(s o j )\u8868\u793a\u53e5\u5b50so j \u7684\u7f16\u7801\u3002 e j = LST M (Enc(s o j , e j\u22121 ), j = (1, \u2022 \u2022 \u2022 , n) (6) 2.2.2 \u89e3\u7801\u7aef \u6307\u9488\u7f51\u7edc\u7684\u89e3\u7801\u5668\u6a21\u578b\u53ef\u4ee5\u8868\u793a\u4e3a\u5f0f(7)\uff0c\u5176\u4e2d\uff0cEnc(s o i )\u8868\u793a\u53e5\u5b50s o i \u7684\u7f16\u7801\u3002 d i = LST M (Enc(s o i , d i\u22121 ), i = (1, \u2022 \u2022 \u2022 , n)", | |
| "eq_num": "(7)" | |
| } | |
| ], | |
| "section": "", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "2.3 \u53e5\u5b50\u987a\u5e8f\u6982\u7387 \u6211\u4eec\u5c06\u53e5\u5b50\u96c6\u7684\u987a\u5e8f\u8868\u793a\u4e3a\uff1aP (o|s)\uff0c\u5c06\u6700\u4f73\u53e5\u5b50\u987a\u5e8f\u8868\u793a\u4e3a\u00f4:", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "EQUATION", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [ | |
| { | |
| "start": 0, | |
| "end": 8, | |
| "text": "EQUATION", | |
| "ref_id": "EQREF", | |
| "raw_str": "o = argmax o P (o|s) (8) \u627e\u5230\u53e5\u5b50\u96c6s\u7684\u6700\u4f73\u987a\u5e8f\u662f\u4e00\u4e2aNP\u95ee\u9898\uff0c\u6709\u4e24\u79cd\u7b56\u7565\u53ef\u4ee5\u7528\u6765\u89e3\u51b3\u8fd9\u4e2a\u95ee\u9898\uff1a\u8d2a\u5fc3\u7b97\u6cd5\u548c \u96c6\u675f\u641c\u7d22\u7b97\u6cd5\u3002 2.3.1 \u8d2a\u5fc3\u7b97\u6cd5 \u8d2a\u5fc3\u7b97\u6cd5(Greedy Algorithm)\u7684\u601d\u60f3\u662f\u6307\uff0c\u5728\u5bf9\u95ee\u9898\u6c42\u89e3\u65f6\uff0c\u603b\u662f\u505a\u51fa\u5728\u5f53\u524d\u770b\u6765\u662f\u6700 \u597d\u7684\u9009\u62e9\uff0c\u4e5f\u5c31\u662f\u8bf4\uff0c\u4e0d\u4ece\u6574\u4f53\u6700\u4f18\u4e0a\u52a0\u4ee5\u8003\u8651\uff0c\u5b83\u6240\u505a\u51fa\u7684\u9009\u62e9\u662f\u5728\u67d0\u79cd\u610f\u4e49\u4e0a\u7684\u5c40\u90e8\u6700\u4f18 \u89e3\u3002\u5728\u6307\u9488\u7f51\u7edc\u7684\u89e3\u7801\u9636\u6bb5\uff0c\u7528\u8d2a\u5fc3\u7b97\u6cd5\u8868\u793a\u987a\u5e8f\u00f4 =\u00f4 1 , \u2022 \u2022 \u2022 ,\u00f4 n \u7684\u751f\u6210\u8fc7\u7a0b\u53ef\u4ee5\u8868\u793a\u4e3a\u5f0f(9)\u3002 o i = argmax o i P (o i |\u00f4 i\u22121 , \u2022 \u2022 \u2022 ,\u00f4 1 , s) (9) 2.3.2 \u96c6\u675f\u641c\u7d22\u7b97\u6cd5 \u96c6\u675f\u641c\u7d22(Beam Search)\u662f\u4e00\u79cd\u542f\u53d1\u5f0f\u56fe\u641c\u7d22\u7b97\u6cd5\uff0c\u901a\u5e38\u7528\u5728\u56fe\u7684\u89e3\u7a7a\u95f4\u6bd4\u8f83\u5927\u7684\u60c5\u51b5\u4e0b\uff0c \u4e3a\u4e86\u51cf\u5c11\u641c\u7d22\u6240\u5360\u7528\u7684\u7a7a\u95f4\u548c\u65f6\u95f4\uff0c\u5728\u6bcf\u4e00\u6b65\u6df1\u5ea6\u6269\u5c55\u7684\u65f6\u5019\uff0c\u526a\u6389\u4e00\u4e9b\u8d28\u91cf\u6bd4\u8f83\u5dee\u7684\u7ed3\u70b9\uff0c \u4fdd\u7559\u4e0b\u4e00\u4e9b\u8d28\u91cf\u8f83\u9ad8\u7684\u7ed3\u70b9\u3002 \u8fd9\u6837\u51cf\u5c11\u4e86\u7a7a\u95f4\u6d88\u8017\uff0c\u5e76\u63d0\u9ad8\u4e86\u65f6\u95f4\u6548\u7387\u3002 \u5728\u6c42\u89e3\u6700\u4f18\u89e3\u65f6\uff0c\u96c6 \u675f\u641c\u7d22\u7b97\u6cd5\u7684\u6bcf\u4e00\u6b65\u603b\u662f\u4fdd\u7559\u6700\u4f18\u7684b\u4e2a\u5019\u9009\u9879\u3002 \u5bf9\u4e8e\u7b2ct\u6b65\u6765\u8bf4\uff0c\u6bcf\u4e2a\u5019\u9009\u89e3\u53ef\u4ee5\u8868\u793a\u4e3a\u00f4 t 1 = o 1 , \u2022 \u2022 \u2022 ,\u00f4 t \uff0c\u5176\u6982\u7387\u4e3a\u5f0f(10)\u3002\u5176\u4e2d\u6982\u7387\u6700\u9760\u524d\u7684b\u4e2a\u5019\u9009\u9879\u5c06\u4f1a\u5728\u7b2ct\u6b65\u88ab\u4fdd\u7559\u3002 P (\u00f4 t 1 |s) = t i=1 P (\u00f4 i |\u00f4 i\u22121 , \u2022 \u2022 \u2022 ,\u00f4 1 , s)", | |
| "eq_num": "(10)" | |
| } | |
| ], | |
| "section": "", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "3 \u6a21\u578b\u8bad\u7ec3 \u5728\u671d\u9c9c\u8bed\u4e2d\uff0c\u4e00\u90e8\u5206\u5f62\u6001\u7d20\u5728\u53e5\u5b50\u4e2d\u7684\u5199\u6cd5\u4e0e\u539f\u5f62\u4e4b\u95f4\u5b58\u5728\u5dee\u5f02\u3002 \u4f8b\u5982\uff0c (\u5b9e\u9645 \u5199\u6cd5)\u21d2 (\u5f62\u6001\u7d20\u539f\u5f62) \u3002\u53ef\u4ee5\u770b\u5230\uff0c\u5728\u5f62\u6001\u7d20\u5206\u6790\u8fc7\u7a0b\u4e2d\uff0c \" \"\u5f62\u6001\u7d20\u8f6c\u5316\u4e3a \" \" \uff0c \" \"\u8fd9\u662f\u56e0\u4e3a\" \" (\u8868\u793a\u5df2\u7ecf\u505a\u5b8c)\u5c5e\u4e8e\u7f29\u7565\u8bed\uff0c\u5176\u4e2d\u5305\u62ec\u4e86\u8bcd\u5e72\u4fe1\u606f\"\u505a\"\u548c\u65f6\u6001 (Simard et al., 2003) \u4eff\u9020\u751f\u7269\u7684\u89c6 (Duchi et al., 2011) \u7ed3\u5408\u5c0f\u6279\u91cf\u68af\u5ea6\u4e0b\u964d (Turian et al., 2010) (1, 3, 4))\u7684\u503c\u5c31\u662f2\u3002 ", | |
| "cite_spans": [ | |
| { | |
| "start": 132, | |
| "end": 153, | |
| "text": "(Simard et al., 2003)", | |
| "ref_id": "BIBREF15" | |
| }, | |
| { | |
| "start": 161, | |
| "end": 181, | |
| "text": "(Duchi et al., 2011)", | |
| "ref_id": "BIBREF5" | |
| }, | |
| { | |
| "start": 192, | |
| "end": 213, | |
| "text": "(Turian et al., 2010)", | |
| "ref_id": "BIBREF17" | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "\u4fe1\u606f\"\u5df2\u7ecf\" \u3002\u89e3\u51b3\u8fd9\u4e00\u95ee\u9898\u7684\u5e38\u7528\u65b9\u6cd5\u662f\u5229\u7528\u8bed\u6599\u5e93\u5efa\u7acb\u5f62\u6001\u7d20\u53d8\u5f62\u8bcd\u5178\uff0c\u5e76\u5229\u7528\u8bcd\u5178\u5b8c\u6210\u5f62 \u6001\u7d20\u539f\u5f62\u6062\u590d\u3002 \u7136\u800c\u57fa\u4e8e\u8bcd\u5178\u7684\u5f62\u6001\u7d20\u539f\u5f62\u6062\u590d\u65b9\u6cd5\u53d7\u9650\u4e8e\u8bed\u6599\u5e93\u8d28\u91cf\uff0c\u5b58\u5728\u5904\u7406\u4e0d\u597d\u7684\u672a\u767b \u5f55\u8bcd\u7b49\u95ee\u9898\u3002 \u9488\u5bf9\u8fd9\u4e00\u95ee\u9898\uff0c\u672c\u6587\u91c7\u7528\u7ed3\u5408\u8bcd\u6027\u4fe1\u606f\u7684\u591a\u4efb\u52a1seq2seq\u6a21\u578b\u6765\u89e3\u51b3\u8fd9\u4e00\u95ee\u9898\u3002 \u5728 \u671d\u9c9c\u8bed\u4e2d\uff0c\u7531\u7a7a\u683c\u5206\u5f00\u7684\u5355\u5143\u662f\u8bed\u8282\u3002 \u7531\u4e8e\u671d\u9c9c\u8bed\u6240\u6709\u7684\u8bed\u8282\u6570\u91cf\u975e\u5e38\u5e9e\u5927\uff0c\u5b9e\u9a8c\u4e2d\u7528\u5230\u7684\u8bed \u6599\u4e2d\u6709624,655\u4e2a\u4e0d\u540c\u7684\u8bed\u8282\uff0c\u4e0d\u592a\u9002\u5408\u76f4\u63a5\u4f5c\u4e3aseq2seq\u6a21\u578b\u7684\u8f93\u5165\u3002\u672c\u6587\u8003\u8651\u5c06\u4e00\u4e2a\u8bed\u8282\u770b\u4f5c \u4e00\u4e2a\u97f3\u8282\u5e8f\u5217\u3002 \u4f8b\u5982\u8bed\u8282' '\u662f4\u4e2a\u97f3\u8282' ' \uff0c ' ' \uff0c ' ' \uff0c ' '\u7ec4\u6210\uff0c\u53ef\u4ee5\u770b\u4f5c\u4e00 \u4e2a\u97f3\u8282\u5e8f\u5217\u3002 \u540c\u6837\u7684\uff0c\u5f62\u6001\u7d20\u4e5f\u662f\u7531\u97f3\u8282\u7ec4\u6210\u7684\uff0c\u4e5f\u53ef\u4ee5\u770b\u4f5c\u4e00\u4e2a\u97f3\u8282\u5e8f\u5217\u3002 \u672c\u6587\u4e2d\u5b9e\u9a8c\u7528\u5230 \u7684\u8bed\u6599\u4e2d\u7684\u97f3\u8282\u6570\u91cf\u4e3a5,", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "\u5c06\u8f93\u5165\u7684\u97f3\u8282\u5e8f\u5217S : (s 1 , \u2022 \u2022 \u2022 , s N )\u8f6c\u5316\u6210\u9690\u85cf\u72b6\u6001\u5e8f\u5217H : (h 1 , \u2022 \u2022 \u2022 , h N )\uff1b \u6ce8\u610f\u529b\u673a\u5236\u6a21 \u57571(Attention1)\u7684\u4f5c\u7528\u662f\u5c06\u9690\u85cf\u72b6\u6001\u5e8f\u5217H : (h 1 , \u2022 \u2022 \u2022 , h N )\u8f6c\u5316\u6210\u8003\u8651\u4e0a\u4e0b\u6587\u4fe1\u606f\u7684\u9690\u85cf\u72b6\u6001\u5e8f \u5217C 1 : (c 1 1 , \u2022 \u2022 \u2022 , c 1 M 1 )\uff0c\u5e76\u8f93\u5165\u7ed9\u89e3\u7801\u6a21\u5757(Decoder)\uff1b\u89e3\u7801\u6a21\u5757(Decoder)\u7684\u4f5c\u7528\u662f\u5c06\u9690\u85cf\u72b6\u6001\u5e8f \u5217C 1 : (c 1 1 , \u2022 \u2022 \u2022 , c 1 M 1 )\u8f6c\u5316\u6210\u8bcd\u6027\u6807\u8bb0\u5e8f\u5217Z : (z 1 , \u2022 \u2022 \u2022 , z T )\uff1b\u6ce8\u610f\u529b\u673a\u5236\u6a21\u57572(Attention2)\u7684\u4f5c\u7528 \u662f\u8003\u8651\u8bcd\u6027\u4fe1\u606f\u7684\u540c\u65f6\uff0c\u5c06\u9690\u85cf\u72b6\u6001\u5e8f\u5217H : (h 1 , \u2022 \u2022 \u2022 , h N )\u8f6c\u5316\u6210\u8003\u8651\u4e0a\u4e0b\u6587\u4fe1\u606f\u7684\u9690\u85cf\u72b6\u6001\u5e8f \u5217C 2 : (c 2 1 , \u2022 \u2022 \u2022 , c 2 M 2 )\uff0c\u5e76\u8f93\u5165\u5230\u6307\u9488\u7f51\u7edc(Pointer Network)\uff1b\u6307\u9488\u7f51\u7edc\u6a21\u5757(Pointer Network)\u7684 \u4f5c\u7528\u662f\u901a\u8fc7softmax\u51fd\u6570\u5f62\u6210\u6307\u9488\uff0c\u4ece\u8f93\u5165\u97f3\u8282\u5e8f\u5217\u6216\u7ed9\u5b9a\u7684\u97f3\u8282\u8868\u4e2d\u9009\u62e9\u97f3\u8282(\u6216\u7a7a\u683c) \uff0c\u751f \u6210\u5f62\u6001\u7d20\u5e8f\u5217Y : (y 1 , \u2022 \u2022 \u2022 , y L )\u3002 \u7ed9\u5b9a\u8f93\u5165\u5e8f\u5217S : (s 1 , \u2022 \u2022 \u2022 , s N )\uff0cS\u8868\u793a\u7684\u662f\u8f93\u5165\u7684\u8bed\u8282\uff0c\u5c06\u8bed\u8282 \u62c6\u5206\u6210\u97f3\u8282\u7ec4\u6210\u7684\u5e8f\u5217\uff0c\u5176\u4e2ds i \u8868\u793a\u7684\u662f\u7b2ci\u4e2a\u97f3\u8282\uff0c\u8f93\u51fa\u5e8f\u5217\u662fY : (y 1 , \u2022 \u2022 \u2022 , y L )\uff0cY \u8868\u793a\u7684\u662f\u62c6 \u5206\u597d\u7684\u5f62\u6001\u7d20\u5e8f\u5217\uff0c\u5176\u4e2dY i \u8868\u793a\u7684\u662f\u7b2ci\u4e2a\u5f62\u6001\u7d20\u3002 \u7528\u8bad\u7ec3\u597d\u7684\u6a21\u578b\u8fdb\u884c\u5f62\u6001\u7d20\u539f\u5f62\u8f6c\u6362\u62c6\u5206\uff0c\u671d\u9c9c\u8bed\u7684\u6700\u5c0f\u5355\u4f4d\u662f\u5f62\u6001\u7d20\u3002", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "EQUATION", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [ | |
| { | |
| "start": 0, | |
| "end": 8, | |
| "text": "EQUATION", | |
| "ref_id": "EQREF", | |
| "raw_str": "\u77e5\u89c9\u673a\u5236\uff0c \u5305\u542b\u5377\u79ef\u8ba1\u7b97\u4e14\u5177\u6709\u6df1\u5ea6\u7ed3\u6784\u7684\u524d\u9988\u795e\u7ecf\u7f51\u7edc\uff0c \u662f\u6df1\u5ea6\u5b66\u4e60\u7684\u4ee3\u8868\u7b97\u6cd5\u4e4b\u4e00\u3002 \u5c06 \u5305\u542bn w \u4e2a\u5355\u8bcd\u7684\u53e5\u5b50s\u901a\u8fc7\u5377\u79ef\u795e\u7ecf\u7f51\u7edc\u7f16\u7801\u7684\u8fc7\u7a0b\u53ef\u4ee5\u8868\u793a\u4e3a\u516c\u5f0f(11)\uff0c(12)\u3002 \u5176\u4e2dW cov \u2208 R (dl f )d f \u548cb cov \u2208 R d f \u662f\u53ef\u8bad\u7ec3\u7684\u53c2\u6570\uff0c \u5176\u4e2d\u03c6(\u2022)\u662ftanh\u51fd\u6570\u3002 k = 1, \u2022 \u2022 \u2022 , n w \u2212 l f + 1\u3002 \u5176 \u4e2d\u7684l f \u548cd f \u90fd\u662f\u5377\u79ef\u795e\u7ecf\u7f51\u7edc\u6a21\u578b\u4e2d\u7684\u8d85\u53c2\u6570\uff0c \u5206\u522b\u662f\u8fc7\u6ee4\u5668(filter)\u7684\u957f\u5ea6\u548c\u7279\u5f81\u56fe(feature map)\u7684\u4e2a\u6570\u3002 cov k = \u03c6(W T cov (\u2295 l f \u22121 u=0 w k+u ) + b cov )", | |
| "eq_num": "(11" | |
| } | |
| ], | |
| "section": "", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "EQUATION", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [ | |
| { | |
| "start": 0, | |
| "end": 8, | |
| "text": "EQUATION", | |
| "ref_id": "EQREF", | |
| "raw_str": "\u50a8\u5355\u5143c \u2208 R dr \u7531\u4e09\u79cd\u95e8\u63a7\u5236\uff1a\u8f93\u5165\u95e8i \u2208 R dr \u3001\u9057\u5fd8\u95e8f \u2208 R dr \u8f93\u51fa\u95e8o \u2208 R dr \uff0c\u8868\u793a\u4e3a\u516c\u5f0f(13)- (15)\u3002 \u5176\u4e2d\uff0cW g \u2208 R (d+dr)4dr \u548cb g \u2208 R 4dr \u662f\u53ef\u8bad\u7ec3\u7684\u53c2\u6570\uff0cd r \u662f\u8868\u793a\u5b58\u50a8\u5355\u5143\u548c\u95e8\u63a7\u5355\u5143\u7684\u7ef4 \u8ba1\u7b97\u8bed\u8a00\u5b66 \u5ea6\u7684\u4e00\u4e2a\u8d85\u53c2\u6570\u3002t = 1, \u2022 \u2022 \u2022 , n w \u5176\u4e2d\u03c3(\u2022)\u662fsigmoid\u51fd\u6570\uff0c\u03c6(\u2022)\u662ftanh\u51fd\u6570\u3002 \uf8ee \uf8ef \uf8ef \uf8ef \uf8ef \uf8ef \uf8f0 i t o t f t c t \uf8f9 \uf8fa \uf8fa \uf8fa \uf8fa \uf8fa \uf8fb = \uf8ee \uf8ef \uf8ef \uf8ef \uf8ef \uf8ef \uf8f0 \u03c3 \u03c3 \u03c3 \u03c6 \uf8f9 \uf8fa \uf8fa \uf8fa \uf8fa \uf8fa \uf8fb (W T g w t h t\u22121 + b g )", | |
| "eq_num": "(13)" | |
| } | |
| ], | |
| "section": "", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "c t = c t\u22121 f t + c t i t (14) h t = o t \u03c6(c t ) (15) \u6211\u4eec\u5c06\u901a\u8fc7\u957f\u77ed\u65f6\u8bb0\u5fc6\u7f51\u7edc\u7f16\u7801\u7684\u53e5\u5b50\u5411\u91cf\u8868\u793a\u4e3a\uff1a Enc(s) = h nw (16) 3.3 \u76ee\u6807\u51fd\u6570\u8bad\u7ec3 \u8bbe\u6709m\u4e2a\u8bad\u7ec3\u6837\u672c(x i , y i ) m i=1 \uff0cx i \u8868\u793a\u7684\u662f\u4e00\u4e2a\u53e5\u5b50\u96c6\u5408\uff0c\u8fd9\u4e2a\u53e5\u5b50\u96c6\u5408\u6709\u4e00\u4e2a\u552f\u4e00\u7279\u5b9a\u7684 \u6392\u5e8f\u5e8f\u5217y i \uff0c y i \u7684\u53e5\u5b50\u987a\u5e8f\u662f\u6700\u4f18\u987a\u5e8fo * \u3002 \u4e3a\u4e86\u5f97\u5230\u66f4\u591a\u7684\u8bad\u7ec3\u6570\u636e\uff0c \u672c\u6587\u5728\u8bad\u7ec3\u6a21\u578b\u7684\u8fc7 \u7a0b\u65f6\uff0c\u5728\u6bcf\u4e2aepoch\u4e2d\u4e3a\u53e5\u5b50\u96c6\u5408x i \u968f\u673a\u751f\u6210\u65b0\u7684\u6392\u5e8f\u3002 \u76ee\u6807\u51fd\u6570\u53ef\u4ee5\u8868\u793a\u4e3a\u516c\u5f0f(17)\u3002 \u5176\u4e2d\uff0c P (y i |x i ; \u03b8) = P (o * |S = x i ; \u03b8)\uff0c\u03bb\u662f\u6b63\u5219\u9879\u7684\u8d85\u53c2\u6570\u3002 \u03b8\u8868\u793a\u6240\u6709\u53ef\u8bad\u7ec3\u7684\u53c2\u6570\u3002 \u6b64\u5916\uff0c\u672c\u6587\u91c7 \u7528AdaGrad", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "\u4f18\u5316\u7b97\u6cd5\u6765\u8bad\u7ec3\u6a21\u578b\u3002 J(\u03b8) = \u2212 1 m m i=1 logP (y i |x i ; \u03b8) + \u03bb 2 \u03b8 2 2 (17) 4 \u5b9e\u9a8c 4.1 \u6570\u636e\u96c6 \u672c \u6587 \u4ece \u5ef6 \u8fb9 \u65e5 \u62a5 \u671d \u9c9c \u8bed \u7248\u3001 \u4eba \u6c11 \u7f51 \u671d \u9c9c \u8bed \u7248 \u7b49 \u65b0 \u95fb \u7f51 \u7ad9 \u722c \u53d6 \u4e8620000\u7bc7 \u671d \u9c9c \u8bed \u65b0 \u95fb \u4f5c \u4e3a \u8bed \u6599\u3002 \u5c06 \u6bcf \u7bc7 \u65b0 \u95fb \u8fdb \u884c \u8bed \u6bb5 \u5206 \u9694\uff0c \u9009 \u53d6 \u53e5 \u5b50 \u6570 \u76ee \u5927 \u4e8e2\u7684 \u8bed \u6bb5 \u4f5c \u4e3a \u4e00 \u4e2a \u6570 \u636e \u5355 \u5143\uff0c \u5c06 \u6bcf \u4e2a \u6570 \u636e \u5355 \u5143 \u7684 \u53e5 \u5b50 \u8fdb \u884c \u6253 \u4e71 \u7f16 \u53f7\u3002 \u4f8b \u5982 \u5c06 \u8bed \u6bb5[s1\uff0c s2\uff0c s3\uff0c s4]\u7f16 \u7801 \u4e3a[4,1,2,3]\uff0c \u7136 \u540e \u518d \u5bf9 \u8be5 \u8bed \u6bb5 \u7f16 \u7801 \u968f \u673a \u6253 \u4e71 \u4e3a[3,2,4,1]\u3002 \u8fd9 \u6837 \u6211 \u4eec \u5c31 \u5f97 \u5230 \u4e00 \u4e2a \u8bad \u7ec3 \u6837 \u672c([\u53e51,\u53e52,\u53e53,\u53e54], [4,1,2,3], [3,2,4,1])\uff0c \u7b2c \u4e00 \u9879 \u4e3a \u987a \u5e8f \u53e5 \u5b50 \u96c6 \u5408\uff0c \u7b2c \u4e8c \u9879 \u4e3a \u6b63 \u786e \u987a \u5e8f\uff0c \u7b2c \u4e09 \u9879 \u4e3a \u4e71 \u5e8f \u987a \u5e8f\u3002 \u6309 \u7167 \u4e0a \u8ff0 \u5f62 \u5f0f \u5bf9 \u6240 \u6709 \u6570 \u636e \u5355 \u5143 \u8fdb \u884c \u7f16 \u7801 \u518d \u6253 \u4e71\uff0c \u5f97 \u5230 \u6837 \u672c \u96c6 \u5408\u3002 \u5bf9 \u8fd9 \u4e9b \u6837 \u672c \u96c6 \u5408 \u8fdb \u884c \u8bad \u7ec3 \u96c6\u3001 \u9a8c \u8bc1 \u96c6 \u548c \u6d4b \u8bd5 \u96c6 \u7684 \u5212 \u5206\u3002 \u5212 \u5206 \u7ed3 \u679c \u5982 \u88681\u6240", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "EQUATION", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [ | |
| { | |
| "start": 0, | |
| "end": 8, | |
| "text": "EQUATION", | |
| "ref_id": "EQREF", | |
| "raw_str": "\u5408\u7684\u5927\u5c0f\u3002 P = 1 m i=1 m |S(\u00f4 i ) \u2229 S(o * i )| |S(\u00f4 i )| (18) R = 1 m i=1 m |S(\u00f4 i ) \u2229 S(o * i )| |S(\u00f4 * i )| (19) F = 2 * P * R P + R", | |
| "eq_num": "(" | |
| } | |
| ], | |
| "section": "", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "EQUATION", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [ | |
| { | |
| "start": 0, | |
| "end": 8, | |
| "text": "EQUATION", | |
| "ref_id": "EQREF", | |
| "raw_str": "P = 1 m m i=1 |L(\u00f4 i , o * i )| |\u00f4 i | (21) P = 1 m m i=1 |L(\u00f4 i , o * i )| |o * i | (22) F = 2 * P * R P + R (23) 4.3.3 \u6700\u4f73\u5339\u914d\u6bd4\u6cd5 \u6700\u4f73\u5339\u914d\u6bd4\u6cd5(Perfect match ratio, PMR)\u8ba1\u7b97\u7684\u662f\u786e\u5207\u7684\u5339\u914d\u9879\u7684\u6bd4\u4f8b\uff0c \u5982\u516c\u5f0f(24)\uff0c (25)\u6240\u793a\u3002\u5176\u4e2d\uff0cP (\u2022)\u8868\u793a\u00f4 i \u548co * i \u7684\u6700\u4f73\u5339\u914d\u5b50\u5e8f\u5217\u7684\u957f\u5ea6\u3002 P M R = 1 m m i=1 1{\u00f4 i = o * i }", | |
| "eq_num": "(24)" | |
| } | |
| ], | |
| "section": "", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "EQUATION", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [ | |
| { | |
| "start": 0, | |
| "end": 8, | |
| "text": "EQUATION", | |
| "ref_id": "EQREF", | |
| "raw_str": "{\u00f4 i = o * i } = P (\u00f4 i \u2229 o * i ) o i", | |
| "eq_num": "(25" | |
| } | |
| ], | |
| "section": "", | |
| "sec_num": null | |
| } | |
| ], | |
| "back_matter": [], | |
| "bib_entries": { | |
| "BIBREF0": { | |
| "ref_id": "b0", | |
| "title": "Tied Multitask Learning for Neural Speech Translation", | |
| "authors": [ | |
| { | |
| "first": "Antonios", | |
| "middle": [], | |
| "last": "Anastasopoulos", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "David", | |
| "middle": [], | |
| "last": "Chiang", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2018, | |
| "venue": "Proceedings of NAACL-HLT", | |
| "volume": "2", | |
| "issue": "", | |
| "pages": "82--91", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Antonios Anastasopoulos and David Chiang. 2018. Tied Multitask Learning for Neural Speech Trans- lation. Association for Computational Linguistics. Proceedings of NAACL-HLT 2, New Orleans, Louisiana, 82-91.", | |
| "links": null | |
| }, | |
| "BIBREF1": { | |
| "ref_id": "b1", | |
| "title": "On the study of Korean natural language processing", | |
| "authors": [ | |
| { | |
| "first": "Yude", | |
| "middle": [], | |
| "last": "Bi", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2011, | |
| "venue": "Chinese", | |
| "volume": "25", | |
| "issue": "", | |
| "pages": "166--169", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Yude Bi. 2011. On the study of Korean natural language processing. Journal of Chinese Information Processing, 25(6):166-169. In Chinese.", | |
| "links": null | |
| }, | |
| "BIBREF2": { | |
| "ref_id": "b2", | |
| "title": "Enriching word vectors with subword information", | |
| "authors": [ | |
| { | |
| "first": "Piotr", | |
| "middle": [], | |
| "last": "Bojanowski", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Edouard", | |
| "middle": [], | |
| "last": "Grave", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Armand", | |
| "middle": [], | |
| "last": "Joulin", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Tomas", | |
| "middle": [], | |
| "last": "Mikolov", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2017, | |
| "venue": "Transactions of the Association for Computational Linguistics", | |
| "volume": "5", | |
| "issue": "", | |
| "pages": "135--146", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Piotr Bojanowski, Edouard Grave, Armand Joulin and Tomas Mikolov. 2017. Enriching word vectors with subword information. Transactions of the Association for Computational Linguistics, 5:135- 146.", | |
| "links": null | |
| }, | |
| "BIBREF3": { | |
| "ref_id": "b3", | |
| "title": "Chris Tar, Yun-Hsuan Sung", | |
| "authors": [ | |
| { | |
| "first": "Daniel", | |
| "middle": [], | |
| "last": "Cer", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Yinfei", | |
| "middle": [], | |
| "last": "Yang", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Sheng-Yi", | |
| "middle": [], | |
| "last": "Kong", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Nan", | |
| "middle": [], | |
| "last": "Hua", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Nicole", | |
| "middle": [], | |
| "last": "Limtiaco", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Rhomni", | |
| "middle": [], | |
| "last": "St", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Noah", | |
| "middle": [], | |
| "last": "John", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Mario", | |
| "middle": [], | |
| "last": "Constant", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Steve", | |
| "middle": [], | |
| "last": "Guajardo-Cespedes", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "", | |
| "middle": [], | |
| "last": "Yuan", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2018, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Daniel Cer, Yinfei Yang, Sheng-yi Kong, Nan Hua, Nicole Limtiaco, Rhomni St. John, Noah Constant, Mario Guajardo-Cespedes, Steve Yuan, Chris Tar, Yun-Hsuan Sung, Brian Strope, Ray Kurzweil. 2018. Universal Sentence Encoder. ArXiv. abs/1803.11175.", | |
| "links": null | |
| }, | |
| "BIBREF4": { | |
| "ref_id": "b4", | |
| "title": "Neural sentence ordering", | |
| "authors": [ | |
| { | |
| "first": "Xinchi", | |
| "middle": [], | |
| "last": "Chen", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Xipeng", | |
| "middle": [], | |
| "last": "Qiu", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Xuanjing", | |
| "middle": [], | |
| "last": "Huang", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2016, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Xinchi Chen, Xipeng Qiu and Xuanjing Huang. 2016. Neural sentence ordering. ArXiv. abs/1607.06952. \u8ba1\u7b97\u8bed\u8a00\u5b66", | |
| "links": null | |
| }, | |
| "BIBREF5": { | |
| "ref_id": "b5", | |
| "title": "Adaptive Subgradient Methods for Online Learning and Stochastic Optimization", | |
| "authors": [ | |
| { | |
| "first": "John", | |
| "middle": [], | |
| "last": "Duchi", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Elad", | |
| "middle": [], | |
| "last": "Hazan", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Yorma", | |
| "middle": [], | |
| "last": "Singer", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2011, | |
| "venue": "Journal of Machine Learning Research", | |
| "volume": "12", | |
| "issue": "", | |
| "pages": "2121--2159", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "John Duchi, Elad Hazan and Yorma Singer. 2011. Adaptive Subgradient Methods for Online Learning and Stochastic Optimization. Journal of Machine Learning Research, 12:2121-2159.", | |
| "links": null | |
| }, | |
| "BIBREF6": { | |
| "ref_id": "b6", | |
| "title": "Neural sentence ordering", | |
| "authors": [ | |
| { | |
| "first": "Jingjing", | |
| "middle": [], | |
| "last": "Gong", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Xinchi", | |
| "middle": [], | |
| "last": "Chen", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Xipeng", | |
| "middle": [], | |
| "last": "Qiu", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2016, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Jingjing Gong, Xinchi Chen and Xipeng Qiu. 2016. Neural sentence ordering. ArXiv. abs/1611.04953.", | |
| "links": null | |
| }, | |
| "BIBREF7": { | |
| "ref_id": "b7", | |
| "title": "The application of latent semantic analysis clustering algorithm in abstract sentence ordering. Application Research of Computers", | |
| "authors": [ | |
| { | |
| "first": "Hongjian", | |
| "middle": [], | |
| "last": "Guo", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Bing", | |
| "middle": [], | |
| "last": "Huang", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2013, | |
| "venue": "Chinese", | |
| "volume": "30", | |
| "issue": "", | |
| "pages": "3299--3301", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Hongjian Guo and Bing Huang. 2013. The application of latent semantic analysis clustering algorithm in abstract sentence ordering. Application Research of Computers, 30(11):3299-3301. In Chinese.", | |
| "links": null | |
| }, | |
| "BIBREF8": { | |
| "ref_id": "b8", | |
| "title": "Long Short-Term Memory", | |
| "authors": [ | |
| { | |
| "first": "Sepp", | |
| "middle": [], | |
| "last": "Hochreiter", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "J\u00fcrgen", | |
| "middle": [], | |
| "last": "Schmidhuber", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1997, | |
| "venue": "Neural Computation", | |
| "volume": "9", | |
| "issue": "8", | |
| "pages": "1735--1780", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Sepp Hochreiter and J\u00fcrgen Schmidhuber. 1997. Long Short-Term Memory. Neural Computation, 9(8):1735-1780.", | |
| "links": null | |
| }, | |
| "BIBREF9": { | |
| "ref_id": "b9", | |
| "title": "A method of sentence ordering based on neural network model", | |
| "authors": [ | |
| { | |
| "first": "Shize", | |
| "middle": [], | |
| "last": "Kang", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Hong", | |
| "middle": [], | |
| "last": "Ma", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Ruiyang", | |
| "middle": [], | |
| "last": "Huang", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2016, | |
| "venue": "Chinese", | |
| "volume": "30", | |
| "issue": "", | |
| "pages": "195--202", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Shize Kang, Hong Ma and Ruiyang Huang. 2016. A method of sentence ordering based on neural network model. Journal of Chinese Information Processing, 30(5):195-202. In Chinese.", | |
| "links": null | |
| }, | |
| "BIBREF10": { | |
| "ref_id": "b10", | |
| "title": "Character-aware neural language models", | |
| "authors": [ | |
| { | |
| "first": "Yoon", | |
| "middle": [], | |
| "last": "Kim", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Yacine", | |
| "middle": [], | |
| "last": "Jernite", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "David", | |
| "middle": [], | |
| "last": "Sonta", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Alexander", | |
| "middle": [ | |
| "M" | |
| ], | |
| "last": "Rush", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2016, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Yoon Kim, Yacine Jernite, David Sonta and Alexander M. Rush 2016. Character-aware neural language models. ArXiv. abs/1508.06615.", | |
| "links": null | |
| }, | |
| "BIBREF11": { | |
| "ref_id": "b11", | |
| "title": "Sentence ordering using recurrent nerual networks", | |
| "authors": [ | |
| { | |
| "first": "Lajanugen", | |
| "middle": [], | |
| "last": "Logeswaran", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Honglak", | |
| "middle": [], | |
| "last": "Lee", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Dragomir", | |
| "middle": [], | |
| "last": "Radev", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2016, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Lajanugen Logeswaran, Honglak Lee and Dragomir Radev. 2016. Sentence ordering using recurrent nerual networks. ArXiv. abs/1611.02654.", | |
| "links": null | |
| }, | |
| "BIBREF12": { | |
| "ref_id": "b12", | |
| "title": "Efficient estimation of word representations in vector space", | |
| "authors": [ | |
| { | |
| "first": "Tomas", | |
| "middle": [], | |
| "last": "Mikolov", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Kai", | |
| "middle": [], | |
| "last": "Chen", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Greg", | |
| "middle": [], | |
| "last": "Corrado", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Jeffrey", | |
| "middle": [], | |
| "last": "Dean", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2013, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Tomas Mikolov, Kai Chen, Greg Corrado and Jeffrey Dean. 2013. Efficient estimation of word represen- tations in vector space. ArXiv. abs/1301.3781.", | |
| "links": null | |
| }, | |
| "BIBREF13": { | |
| "ref_id": "b13", | |
| "title": "Abstractive Text Summarization using Sequence-to-sequence RNNs and Beyond", | |
| "authors": [ | |
| { | |
| "first": "Ramesh", | |
| "middle": [], | |
| "last": "Nallapati", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Bowen", | |
| "middle": [], | |
| "last": "Zhou", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "\u00c7", | |
| "middle": [], | |
| "last": "Cicero Dos Santos", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Bing", | |
| "middle": [], | |
| "last": "Aglar Gul\u00e7ehre", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "", | |
| "middle": [], | |
| "last": "Xiang", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2018, | |
| "venue": "Proceedings of The 20th SIGNLL Conference on Computational Natural Language Learning", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "280--290", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Ramesh Nallapati, Bowen Zhou, Cicero dos Santos, \u00c7 aglar Gul\u00e7ehre and Bing Xiang. 2018. Abstractive Text Summarization using Sequence-to-sequence RNNs and Beyond. Association for Computational Linguistics. Proceedings of The 20th SIGNLL Conference on Computational Natural Language Learning, Berlin, Germany, 280-290.", | |
| "links": null | |
| }, | |
| "BIBREF14": { | |
| "ref_id": "b14", | |
| "title": "Subword-level Word Vector Representations for Korean", | |
| "authors": [ | |
| { | |
| "first": "Sungjoon", | |
| "middle": [], | |
| "last": "Park", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Jeongmin", | |
| "middle": [], | |
| "last": "Byun", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Sion", | |
| "middle": [], | |
| "last": "Baek", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Yongseok", | |
| "middle": [], | |
| "last": "Cho", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Alice", | |
| "middle": [], | |
| "last": "Oh", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2018, | |
| "venue": "Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics", | |
| "volume": "1", | |
| "issue": "", | |
| "pages": "2429--2438", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Sungjoon Park, Jeongmin Byun, Sion Baek, Yongseok Cho and Alice Oh. 2018. Subword-level Word Vector Representations for Korean. Association for Computational Linguistics. Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), Melbourne, Australia, 2429-2438.", | |
| "links": null | |
| }, | |
| "BIBREF15": { | |
| "ref_id": "b15", | |
| "title": "Best Practices for Convolutional Neural Networks", | |
| "authors": [ | |
| { | |
| "first": "Patrice", | |
| "middle": [ | |
| "Y" | |
| ], | |
| "last": "Simard", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Dave", | |
| "middle": [], | |
| "last": "Steinkraus", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "John", | |
| "middle": [ | |
| "C" | |
| ], | |
| "last": "Platt", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2003, | |
| "venue": "Seventh International Conference on Document Analysis and Recognition", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "958--963", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Patrice Y. Simard, Dave Steinkraus and John C. Platt. 2003. Best Practices for Convolutional Neural Networks. Seventh International Conference on Document Analysis and Recognition, Edinburgh, UK, 958-963.", | |
| "links": null | |
| }, | |
| "BIBREF16": { | |
| "ref_id": "b16", | |
| "title": "Korean Morphological Analysis with Tied Sequence-to-Sequence Multi-Task Model", | |
| "authors": [ | |
| { | |
| "first": "Hyun Je", | |
| "middle": [], | |
| "last": "Song", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Seong Bae", | |
| "middle": [], | |
| "last": "Park", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2019, | |
| "venue": "Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "1436--1441", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Hyun Je Song and Seong Bae Park. 2019. Korean Morphological Analysis with Tied Sequence-to- Sequence Multi-Task Model . Association for Computational Linguistics. Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing, Hong Kong, China, 1436-1441.", | |
| "links": null | |
| }, | |
| "BIBREF17": { | |
| "ref_id": "b17", | |
| "title": "Word representations: a simple and general method for semi-supervised learning", | |
| "authors": [ | |
| { | |
| "first": "Joseph", | |
| "middle": [], | |
| "last": "Turian", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Lev-Arie", | |
| "middle": [], | |
| "last": "Ratinov", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Yoshua", | |
| "middle": [], | |
| "last": "Bengio", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2010, | |
| "venue": "Proceedings of the 48th Annual Meeting of the Association for Computational Linguistics", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "384--394", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Joseph Turian, Lev-Arie Ratinov and Yoshua Bengio. 2010. Word representations: a simple and general method for semi-supervised learning. Association for Computational Linguistics. Proceedings of the 48th Annual Meeting of the Association for Computational Linguistics, Uppsala, Sweden, 384-394.", | |
| "links": null | |
| }, | |
| "BIBREF18": { | |
| "ref_id": "b18", | |
| "title": "Pointer Networks", | |
| "authors": [ | |
| { | |
| "first": "Oriol", | |
| "middle": [], | |
| "last": "Vinyals", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Meire", | |
| "middle": [], | |
| "last": "Fortunato", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Navdeep", | |
| "middle": [], | |
| "last": "Jaitly", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2015, | |
| "venue": "Advances in Neural Information Processing Systems", | |
| "volume": "28", | |
| "issue": "", | |
| "pages": "2692--2700", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Oriol Vinyals, Meire Fortunato and Navdeep Jaitly. 2015. Pointer Networks. Advances in Neural Information Processing Systems 28, 2692-2700.", | |
| "links": null | |
| }, | |
| "BIBREF19": { | |
| "ref_id": "b19", | |
| "title": "Research on the strategy of sentence ordering based on time information in multi document summarization", | |
| "authors": [ | |
| { | |
| "first": "Yongdong", | |
| "middle": [], | |
| "last": "Xu", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Yadong", | |
| "middle": [], | |
| "last": "Wang", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Yang", | |
| "middle": [], | |
| "last": "Liu", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2009, | |
| "venue": "Chinese", | |
| "volume": "23", | |
| "issue": "", | |
| "pages": "27--33", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Yongdong Xu, Yadong Wang and Yang Liu. 2009. Research on the strategy of sentence ordering based on time information in multi document summarization. Journal of Chinese Information Processing, 23(4):27-33. In Chinese.", | |
| "links": null | |
| }, | |
| "BIBREF20": { | |
| "ref_id": "b20", | |
| "title": "Research on sentence ordering based on conditional entropy and context proximity", | |
| "authors": [ | |
| { | |
| "first": "Tao", | |
| "middle": [], | |
| "last": "Xue", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Heng", | |
| "middle": [], | |
| "last": "Wang", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2017, | |
| "venue": "Chinese", | |
| "volume": "34", | |
| "issue": "", | |
| "pages": "2680--2684", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Tao Xue and Heng Wang. 2017. Research on sentence ordering based on conditional entropy and context proximity. Application Research of Computers, 34(9):2680-2684. In Chinese.", | |
| "links": null | |
| }, | |
| "BIBREF21": { | |
| "ref_id": "b21", | |
| "title": "Sentence ordering of multi document abstracts based on cohesion. The academic conference of the 25th anniversary of the Chinese information society", | |
| "authors": [ | |
| { | |
| "first": "Chao", | |
| "middle": [], | |
| "last": "Yao", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Sheng", | |
| "middle": [], | |
| "last": "Li", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Shu", | |
| "middle": [], | |
| "last": "Zhang", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2006, | |
| "venue": "Chinese", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Chao Yao, Sheng Li and Shu Zhang. 2006. Sentence ordering of multi document abstracts based on cohesion. The academic conference of the 25th anniversary of the Chinese information society. In Chinese.", | |
| "links": null | |
| } | |
| }, | |
| "ref_entries": { | |
| "FIGREF0": { | |
| "uris": null, | |
| "type_str": "figure", | |
| "text": "Xu et al., 2009)\uff1b \u59da\u8d85\u63d0\u51fa\u4e86\u4e00\u79cd\u57fa\u4e8e\u5185\u805a\u5ea6\u7684\u591a\u6587\u6863\u6587\u6458\u7684\u53e5\u5b50\u6392\u5e8f\u65b9\u6cd5\uff0c\u901a\u8fc7\u5c06\u76f8\u540c\u8bdd\u9898\u7684\u53e5\u5b50\u805a\u5408\u5230\u4e00\u8d77\uff0c \u907f\u514d\u8bdd\u9898\u4e2d\u65ad\uff0c\u6539\u5584\u6587\u6458\u53ef\u8bfb\u6027(Yao et al., 2006)\uff1b\u859b\u6d9b\u5c06\u6761\u4ef6\u71b5\u5f15\u5165\u5230\u53e5\u5b50\u6392\u5e8f\u5de5\u4f5c\u4e2d\uff0c\u901a\u8fc7 \u5728\u6e90\u6587\u6863\u4e2d\u8ba1\u7b97\u53e5\u5b50\u5bf9\u7684\u8f6c\u79fb\u4fe1\u606f\u91cf\u6765\u8861\u91cf\u53e5\u5b50\u7684\u5173\u8054\u7a0b\u5ea6\uff0c\u540c\u65f6\u63d0\u51fa\u4e0a\u4e0b\u6587\u5bf9\u6bd4\u7b97\u6cd5\u6765\u52a0\u5f3a \u53e5\u5b50\u90bb\u8fd1\u5ea6\u5b66\u4e60\u7684\u51c6\u786e\u6027(Xue and Wang, 2017)\uff1b\u90ed\u7ea2\u5efa\u5c06\u6f5c\u5728\u8bed\u4e49\u5206\u6790\u805a\u7c7b\u7b97\u6cd5\u5f15\u5165\u6587\u6458\u53e5 \u5b50\u6392\u5e8f\u8fc7\u7a0b\u4e2d\uff0c\u5c06\u8bdd\u9898\u805a\u7c7b\u4e4b\u540e\u91c7\u7528\u6a21\u677f\u5bf9\u6587\u6458\u53e5\u5b50\u8fdb\u884c\u4e24\u8d9f\u6392\u5e8f(Guo and Huang, 2013)\u2026\u2026 \u4f46\u662f\uff0c\u968f\u7740\u5927\u6570\u636e\u3001\u4e91\u8ba1\u7b97\u7b49\u6280\u672f\u7684\u53d1\u5c55\uff0c\u6df1\u5ea6\u5b66\u4e60\u65b9\u6cd5\u5728\u81ea\u7136\u8bed\u8a00\u5904\u7406\u4efb\u52a1\u4e2d\u5e7f\u6cdb\u5e94\u7528\uff0c \u5f88\u591a\u6df1\u5ea6\u5b66\u4e60\u65b9\u6cd5\u88ab\u5f15\u5165\u5230\u53e5\u5b50\u6392\u5e8f\u4e2d\u3002 \u5eb7\u4e16\u6cfd\u5229\u7528\u795e\u7ecf\u7f51\u7edc\u5c06\u51e0\u79cd\u524d\u4eba\u63d0\u51fa\u7684\u53e5\u5b50\u6392\u5e8f\u65b9\u6cd5 \u878d\u5408\uff0c\u5e76\u5728\u6b64\u57fa\u7840\u4e0a\u63d0\u51fa\u4e86\u4e00\u79cd\u57fa\u4e8e\u9a6c\u5c14\u79d1\u592b\u968f\u673a\u6e38\u8d70\u6a21\u578b\u7684\u53e5\u5b50\u6392\u5e8f\u7b97\u6cd5(Kang et al., 2016)\u3002 Chen\u5c1d\u8bd5\u4e86\u57fa\u4e8e\u5377\u79ef\u795e\u7ecf\u7f51\u7edc(convolutional neural netw-orks, CNNs)\uff0c\u957f\u77ed\u671f\u8bb0\u5fc6\u7f51\u7edc(long short-term memory network, LSTM)\u7684\u53e5\u5b50\u6392\u5e8f\u65b9\u6cd5\uff0c\u4f7f\u7528CNN\u3001 LSTM\u7b49\u6a21\u578b\u5224\u65ad\u53e5\u5b50\u7684\u524d \u540e\u53e5\u5173\u7cfb\uff0c\u5e76\u5229\u7528\u96c6\u675f\u641c\u7d22\u7b97\u6cd5\u6c42\u89e3\u53e5\u5b50\u7684\u6700\u4f18\u6392\u5e8f(Chen et al., 2016)\u3002 Logeswaran\u63d0\u51fa\u4e86 \u4e00\u79cd\u57fa\u4e8e\u5faa\u73af\u795e\u7ecf\u7f51\u7edc\u7684\u53e5\u5b50\u6392\u5e8f\u65b9\u6cd5\uff0c\u901a\u8fc7\u5224\u65ad\u53e5\u5b50\u5728\u6bcf\u4e2a\u4f4d\u7f6e\u7684\u53ef\u80fd\u6027\uff0c\u6c42\u5f97\u6700\u4f18\u6392\u5e8f\u7ed3 \u679c(Logeswaran et al., 2016)\u3002 Gong\u63d0\u51fa\u4e86\u4e00\u79cd\u57fa\u4e8e\u7aef\u5230\u7aef\u7684\u6307\u9488\u7f51\u7edc\u7684\u53e5\u5b50\u6392\u5e8f\u65b9\u6cd5\uff0c\u901a\u8fc7\u7aef \u5230\u7aef\u7684\u6307\u9488\u7f51\u7edc\u5224\u65ad\u6bcf\u4e2a\u4f4d\u7f6e\u4e0a\u7684\u53e5\u5b50\u7684\u53ef\u80fd\u6027\uff0c\u6c42\u5f97\u8f83\u4f18\u6392\u5e8f\u7ed3\u679c(Gong et al., 2016)\u3002 \u672c\u6587\u7684\u4e3b\u8981\u8d21\u732e\u5982\u4e0b\uff1a 1)\u5bf9\u671d\u9c9c\u8bed\u53e5\u5b50\u6392\u5e8f\u95ee\u9898\u8fdb\u884c\u7814\u7a76\uff1b 2)\u5c06\u540c\u5f62\u5f02\u4e49\u8bcd\u4fe1\u606f\u878d\u5165\u5230\u671d\u9c9c\u8bed\u8bcd\u5411\u91cf\u7684\u8bad\u7ec3\uff1b \u00a92020 \u4e2d\u56fd\u8ba1\u7b97\u8bed\u8a00\u5b66\u5927\u4f1a \u6839\u636e\u300aCreative Commons Attribution 4.0 International License\u300b\u8bb8\u53ef\u51fa\u7248", | |
| "num": null | |
| }, | |
| "FIGREF1": { | |
| "uris": null, | |
| "type_str": "figure", | |
| "text": "\u6a21\u578b\u67094\u4e2a\u90e8\u5206\uff1a \u7f16\u7801\u6a21\u5757\u3001 \u89e3\u7801\u6a21\u5757\u3001 \u6ce8\u610f\u529b\u673a\u5236\u6a21\u57571\u3001 \u6ce8\u610f\u529b\u673a\u5236\u6a21\u57572\u3001 \u6307\u9488\u7f51 \u7edc\u6a21\u5757(Nallapati et al., 2016)\u3002 \u56fe4\u6240\u793a\u7684\u662f\u6a21\u578b\u7684\u6574\u4f53\u6846\u67b6\u3002 \u7f16\u7801\u6a21\u5757(Encoder)\u7684\u4f5c\u7528\u662f", | |
| "num": null | |
| }, | |
| "FIGREF2": { | |
| "uris": null, | |
| "type_str": "figure", | |
| "text": "sequence ratio, LSR)\u8ba1\u7b97\u6700\u957f\u6b63\u786e\u5b50\u5e8f\u5217\u7684\u6bd4(\u4e0d\u9700\u8981\u8fde\u7eed\u6027\uff0c\u8d8a \u9ad8\u8d8a\u597d) \u3002 \u6700\u957f\u5e8f\u5217\u6bd4\u6cd5\u53ef\u4ee5\u8868\u793a\u4e3a\u4e09\u4e2a\u5206\u6570\uff1a\u7cbe\u786e\u503cP\u3001 \u53ec\u56de\u7387R\u548cF\u503c\uff0c\u5982\u516c\u5f0f(21)-(23)\u6240 \u793a\u3002 \u5176\u4e2d\uff0c \u51fd\u6570L(\u2022)\u8868\u793a\u7684\u662f\u6700\u957f\u6b63\u786e\u5b50\u5e8f\u5217\u4e2d\u5143\u7d20\u7684\u4e2a\u6570\u3002 \u90a3\u4e48\uff0c L(\u00f4 = (2, 3, 1, 4), o * =", | |
| "num": null | |
| } | |
| } | |
| } | |
| } |