| { |
| "paper_id": "2019", |
| "header": { |
| "generated_with": "S2ORC 1.0.0", |
| "date_generated": "2023-01-19T14:54:48.639152Z" |
| }, |
| "title": "Novel Sequence Generation Framework for Chinese Abstractive Summarization", |
| "authors": [ |
| { |
| "first": "Chin-Yueh", |
| "middle": [], |
| "last": "\u7c21\u9756\u5cb3", |
| "suffix": "", |
| "affiliation": { |
| "laboratory": "", |
| "institution": "National Taiwan University of Science and Technology", |
| "location": {} |
| }, |
| "email": "" |
| }, |
| { |
| "first": "", |
| "middle": [], |
| "last": "Chien", |
| "suffix": "", |
| "affiliation": { |
| "laboratory": "", |
| "institution": "National Taiwan University of Science and Technology", |
| "location": {} |
| }, |
| "email": "" |
| }, |
| { |
| "first": "Kuan-Yu", |
| "middle": [], |
| "last": "\u9673\u51a0\u5b87", |
| "suffix": "", |
| "affiliation": { |
| "laboratory": "", |
| "institution": "National Taiwan University of Science and Technology", |
| "location": {} |
| }, |
| "email": "" |
| }, |
| { |
| "first": "", |
| "middle": [], |
| "last": "Chen", |
| "suffix": "", |
| "affiliation": { |
| "laboratory": "", |
| "institution": "National Taiwan University of Science and Technology", |
| "location": {} |
| }, |
| "email": "kychen@mail.ntust.edu.tw" |
| } |
| ], |
| "year": "", |
| "venue": null, |
| "identifiers": {}, |
| "abstract": "", |
| "pdf_parse": { |
| "paper_id": "2019", |
| "_pdf_hash": "", |
| "abstract": [], |
| "body_text": [ |
| { |
| "text": "EQUATION", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [ |
| { |
| "start": 0, |
| "end": 8, |
| "text": "EQUATION", |
| "ref_id": "EQREF", |
| "raw_str": "= ( * * + + + ) \u5f0f(5) ( ) = (1 \u2212 ) ( ) + ( ) \u5f0f(6)", |
| "eq_num": "(\u4e09)" |
| } |
| ], |
| "section": "", |
| "sec_num": null |
| }, |
| { |
| "text": "EQUATION", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [ |
| { |
| "start": 0, |
| "end": 8, |
| "text": "EQUATION", |
| "ref_id": "EQREF", |
| "raw_str": "= ( \u22121 ( 1 \u22121 ) \u22ba \u221a ) 1 \u22121 \u5f0f(7) * = ( ( 1 ) \u22ba \u221a ) 1 \u5f0f(8) \u8868\u793a\u5411\u91cf\u7684\u7dad\u5ea6\uff0c\u22ba\u8868\u793a\u77e9\u9663\u7684\u8f49\u7f6e\uff0c\u4e26\u4e14 (", |
| "eq_num": "( 1 )" |
| } |
| ], |
| "section": "", |
| "sec_num": null |
| } |
| ], |
| "back_matter": [], |
| "bib_entries": { |
| "BIBREF0": { |
| "ref_id": "b0", |
| "title": "Young Scholar Fellowship Program), and by the Project J367B83100 (ITRI) under the sponsorship of the Ministry of Economic Affairs", |
| "authors": [], |
| "year": null, |
| "venue": "", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "grant MOST 108-2636-E-011-005 (Young Scholar Fellowship Program), and by the Project J367B83100 (ITRI) under the sponsorship of the Ministry of Economic Affairs, Taiwan. \u53c3\u8003\u6587\u737b", |
| "links": null |
| }, |
| "BIBREF1": { |
| "ref_id": "b1", |
| "title": "A Neural Attention Model for Sentence Summarization", |
| "authors": [ |
| { |
| "first": "A", |
| "middle": [ |
| "M" |
| ], |
| "last": "Rush", |
| "suffix": "" |
| }, |
| { |
| "first": "S", |
| "middle": [], |
| "last": "Harvard", |
| "suffix": "" |
| }, |
| { |
| "first": "S", |
| "middle": [], |
| "last": "Chopra", |
| "suffix": "" |
| }, |
| { |
| "first": "J", |
| "middle": [], |
| "last": "Weston", |
| "suffix": "" |
| } |
| ], |
| "year": 2017, |
| "venue": "ACLWeb. Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "A. M. Rush, S. Harvard, S. Chopra, and J. Weston, \"A Neural Attention Model for Sentence Summarization,\" in ACLWeb. Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing, 2017.", |
| "links": null |
| }, |
| "BIBREF2": { |
| "ref_id": "b2", |
| "title": "A deep reinforced model for abstractive summarization", |
| "authors": [ |
| { |
| "first": "R", |
| "middle": [], |
| "last": "Paulus", |
| "suffix": "" |
| }, |
| { |
| "first": "C", |
| "middle": [], |
| "last": "Xiong", |
| "suffix": "" |
| }, |
| { |
| "first": "R", |
| "middle": [], |
| "last": "Socher", |
| "suffix": "" |
| } |
| ], |
| "year": 2017, |
| "venue": "", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": { |
| "arXiv": [ |
| "arXiv:1705.04304" |
| ] |
| }, |
| "num": null, |
| "urls": [], |
| "raw_text": "R. Paulus, C. Xiong, and R. Socher, \"A deep reinforced model for abstractive summarization,\" arXiv:1705.04304, 2017.", |
| "links": null |
| }, |
| "BIBREF3": { |
| "ref_id": "b3", |
| "title": "Abstractive text summarization using sequence-to-sequence rnns and beyond", |
| "authors": [ |
| { |
| "first": "R", |
| "middle": [], |
| "last": "Nallapati", |
| "suffix": "" |
| }, |
| { |
| "first": "B", |
| "middle": [], |
| "last": "Zhou", |
| "suffix": "" |
| }, |
| { |
| "first": "C", |
| "middle": [], |
| "last": "Gulcehre", |
| "suffix": "" |
| }, |
| { |
| "first": "B", |
| "middle": [], |
| "last": "Xiang", |
| "suffix": "" |
| } |
| ], |
| "year": 2016, |
| "venue": "", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "R. Nallapati, B. Zhou, C. Gulcehre, and B. Xiang, \"Abstractive text summarization using sequence-to-sequence rnns and beyond,\" arXiv preprint arXiv:.06023, 2016.", |
| "links": null |
| }, |
| "BIBREF4": { |
| "ref_id": "b4", |
| "title": "Convolutional sequence to sequence learning", |
| "authors": [ |
| { |
| "first": "J", |
| "middle": [], |
| "last": "Gehring", |
| "suffix": "" |
| }, |
| { |
| "first": "M", |
| "middle": [], |
| "last": "Auli", |
| "suffix": "" |
| }, |
| { |
| "first": "D", |
| "middle": [], |
| "last": "Grangier", |
| "suffix": "" |
| }, |
| { |
| "first": "D", |
| "middle": [], |
| "last": "Yarats", |
| "suffix": "" |
| }, |
| { |
| "first": "Y", |
| "middle": [ |
| "N" |
| ], |
| "last": "Dauphin", |
| "suffix": "" |
| } |
| ], |
| "year": 2017, |
| "venue": "Proceedings of the 34th International Conference on Machine Learning", |
| "volume": "70", |
| "issue": "", |
| "pages": "1243--1252", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "J. Gehring, M. Auli, D. Grangier, D. Yarats, and Y. N. Dauphin, \"Convolutional sequence to sequence learning,\" in Proceedings of the 34th International Conference on Machine Learning-Volume 70, 2017, pp. 1243-1252: JMLR. org.", |
| "links": null |
| }, |
| "BIBREF5": { |
| "ref_id": "b5", |
| "title": "Abstractive sentence summarization with attentive recurrent neural networks", |
| "authors": [ |
| { |
| "first": "S", |
| "middle": [], |
| "last": "Chopra", |
| "suffix": "" |
| }, |
| { |
| "first": "M", |
| "middle": [], |
| "last": "Auli", |
| "suffix": "" |
| }, |
| { |
| "first": "A", |
| "middle": [ |
| "M" |
| ], |
| "last": "Rush", |
| "suffix": "" |
| } |
| ], |
| "year": 2016, |
| "venue": "Proceedings of the the North American Chapter of the Association for Computational Linguistics: Human Language Technologies", |
| "volume": "", |
| "issue": "", |
| "pages": "93--98", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "S. Chopra, M. Auli, and A. M. Rush, \"Abstractive sentence summarization with attentive recurrent neural networks,\" in Proceedings of the the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, 2016, pp. 93-98.", |
| "links": null |
| }, |
| "BIBREF6": { |
| "ref_id": "b6", |
| "title": "Attention is all you need", |
| "authors": [ |
| { |
| "first": "A", |
| "middle": [], |
| "last": "Vaswani", |
| "suffix": "" |
| } |
| ], |
| "year": 2017, |
| "venue": "Advances in neural information processing systems", |
| "volume": "", |
| "issue": "", |
| "pages": "5998--6008", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "A. Vaswani et al., \"Attention is all you need,\" in Advances in neural information processing systems, 2017, pp. 5998-6008.", |
| "links": null |
| }, |
| "BIBREF7": { |
| "ref_id": "b7", |
| "title": "Sequence to sequence learning with neural networks", |
| "authors": [ |
| { |
| "first": "I", |
| "middle": [], |
| "last": "Sutskever", |
| "suffix": "" |
| }, |
| { |
| "first": "O", |
| "middle": [], |
| "last": "Vinyals", |
| "suffix": "" |
| } |
| ], |
| "year": 2014, |
| "venue": "Advances in neural information processing systems", |
| "volume": "", |
| "issue": "", |
| "pages": "3104--3112", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "I. Sutskever and O. Vinyals, \"Sequence to sequence learning with neural networks,\" in Advances in neural information processing systems, 2014, pp. 3104-3112.", |
| "links": null |
| }, |
| "BIBREF8": { |
| "ref_id": "b8", |
| "title": "Long short-term memory", |
| "authors": [ |
| { |
| "first": "S", |
| "middle": [], |
| "last": "Hochreiter", |
| "suffix": "" |
| }, |
| { |
| "first": "J", |
| "middle": [ |
| "J N C" |
| ], |
| "last": "Schmidhuber", |
| "suffix": "" |
| } |
| ], |
| "year": 1997, |
| "venue": "", |
| "volume": "9", |
| "issue": "", |
| "pages": "1735--1780", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "S. Hochreiter and J. J. N. c. Schmidhuber, \"Long short-term memory,\" vol. 9, no. 8, pp. 1735-1780, 1997.", |
| "links": null |
| }, |
| "BIBREF9": { |
| "ref_id": "b9", |
| "title": "A neural attention model for abstractive sentence summarization", |
| "authors": [ |
| { |
| "first": "A", |
| "middle": [ |
| "M" |
| ], |
| "last": "Rush", |
| "suffix": "" |
| }, |
| { |
| "first": "S", |
| "middle": [], |
| "last": "Chopra", |
| "suffix": "" |
| }, |
| { |
| "first": "J", |
| "middle": [], |
| "last": "Weston", |
| "suffix": "" |
| } |
| ], |
| "year": 2015, |
| "venue": "", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "A. M. Rush, S. Chopra, and J. Weston, \"A neural attention model for abstractive sentence summarization,\" arXiv:.00685, 2015.", |
| "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.-W", |
| "middle": [], |
| "last": "Chang", |
| "suffix": "" |
| }, |
| { |
| "first": "K", |
| "middle": [], |
| "last": "Lee", |
| "suffix": "" |
| }, |
| { |
| "first": "K", |
| "middle": [], |
| "last": "Toutanova", |
| "suffix": "" |
| } |
| ], |
| "year": 2018, |
| "venue": "", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": { |
| "arXiv": [ |
| "arXiv:1810.04805" |
| ] |
| }, |
| "num": null, |
| "urls": [], |
| "raw_text": "J. Devlin, M.-W. Chang, K. Lee, and K. Toutanova, \"Bert: Pre-training of deep bidirectional transformers for language understanding,\" arXiv:1810.04805, 2018.", |
| "links": null |
| }, |
| "BIBREF11": { |
| "ref_id": "b11", |
| "title": "Sentence Centrality Revisited for Unsupervised Summarization", |
| "authors": [ |
| { |
| "first": "H", |
| "middle": [], |
| "last": "Zheng", |
| "suffix": "" |
| }, |
| { |
| "first": "M", |
| "middle": [], |
| "last": "Lapata", |
| "suffix": "" |
| } |
| ], |
| "year": 2019, |
| "venue": "", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "H. Zheng and M. Lapata, \"Sentence Centrality Revisited for Unsupervised Summarization,\" arXiv:.03508, 2019.", |
| "links": null |
| }, |
| "BIBREF12": { |
| "ref_id": "b12", |
| "title": "Rouge: A package for automatic evaluation of summaries", |
| "authors": [ |
| { |
| "first": "C.-Y.", |
| "middle": [], |
| "last": "Lin", |
| "suffix": "" |
| } |
| ], |
| "year": 2004, |
| "venue": "Text summarization branches out", |
| "volume": "", |
| "issue": "", |
| "pages": "74--81", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "C.-Y. Lin, \"Rouge: A package for automatic evaluation of summaries,\" in Text summarization branches out, 2004, pp. 74-81.", |
| "links": null |
| }, |
| "BIBREF13": { |
| "ref_id": "b13", |
| "title": "XLNet: Generalized Autoregressive Pretraining for Language Understanding", |
| "authors": [ |
| { |
| "first": "Z", |
| "middle": [], |
| "last": "Yang", |
| "suffix": "" |
| }, |
| { |
| "first": "Z", |
| "middle": [], |
| "last": "Dai", |
| "suffix": "" |
| }, |
| { |
| "first": "Y", |
| "middle": [], |
| "last": "Yang", |
| "suffix": "" |
| }, |
| { |
| "first": "J", |
| "middle": [], |
| "last": "Carbonell", |
| "suffix": "" |
| }, |
| { |
| "first": "R", |
| "middle": [], |
| "last": "Salakhutdinov", |
| "suffix": "" |
| }, |
| { |
| "first": "Q", |
| "middle": [ |
| "V" |
| ], |
| "last": "Le", |
| "suffix": "" |
| } |
| ], |
| "year": 2019, |
| "venue": "", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": { |
| "arXiv": [ |
| "arXiv:1906.08237" |
| ] |
| }, |
| "num": null, |
| "urls": [], |
| "raw_text": "Z. Yang, Z. Dai, Y. Yang, J. Carbonell, R. Salakhutdinov, and Q. V. Le, \"XLNet: Generalized Autoregressive Pretraining for Language Understanding,\" arXiv:1906.08237, 2019.", |
| "links": null |
| } |
| }, |
| "ref_entries": { |
| "TABREF0": { |
| "num": null, |
| "html": null, |
| "text": "\u5168\u4f86\u81ea\u539f\u59cb\u6587\u7ae0\uff0c\u9019\u7a2e\u6458\u8981\u65b9\u5f0f\u53ef\u8aaa\u662f\u66f4\u8cbc\u8fd1\u4eba\u985e\u5e73\u5e38\u64b0\u5beb\u6458\u8981\u7684\u5f62\u5f0f\u3002 \u8fd1\u5e74\u4f86\u81ea\u52d5\u6458\u8981\u7684\u7814\u7a76\u4e2d\uff0c\u5e8f\u5217\u5c0d\u5e8f\u5217\u6a21\u578b\u61c9\u7528\u65bc\u91cd\u5beb\u5f0f\u6458\u8981\u7684\u7814\u7a76[1-5]\uff0c\u5728\u773e\u591a \u8cc7\u6599\u96c6\u4e2d\u9a57\u8b49\u5176\u8c50\u78a9\u7684\u6210\u679c\u3002\u7279\u5225\u662f\u8fd1\u5e74\u63d0\u51fa\u7684\u6307\u91dd\u751f\u6210\u7db2\u8def(Pointer Generator Network, PGN)\uff0c\u5176\u6a5f\u5236\u53ef\u4ee5\u6709\u6548\u89e3\u6c7a\u6587\u7ae0\u4e2d\u5b58\u5728\u975e\u5b57\u5178\u8a5e\u5f59(Out-of-vocabulary)\u7684\u554f\u984c\uff0c\u56e0\u6b64\u88ab \u61c9\u7528\u5728\u91cd\u5beb\u5f0f\u6458\u8981\u4e0a\u3002\u8fd1\u5e74\u7531\u8c37\u6b4c\u63d0\u51fa\u7684 Transformer[6]\u67b6\u69cb\uff0c\u4f7f\u7528\u6ce8\u610f\u529b(Attention)\u6a5f \u5236\uff0c\u53ef\u4ee5\u89e3\u6c7a\u905e\u8ff4\u795e\u7d93\u7db2\u8def\u7684\u9577\u6642\u9593\u5e8f\u5217\u4fe1\u606f\u4e1f\u5931\u554f\u984c\u4e26\u5e73\u884c\u8655\u7406\uff0c\u52a0\u901f\u7db2\u904b\u7b97\u3002\u8c37\u6b4c \u57fa\u65bc Transformer \u67b6\u69cb\u9032\u4e00\u6b65\u5730\u63d0\u51fa BERT\uff0c\u5728\u81ea\u7136\u8a9e\u8a00\u8655\u7406(Natural Language Processing, NLP)\u7684\u5404\u9805\u4efb\u52d9\u4e2d\uff0c\u4f7f\u7528 BERT \u6548\u679c\u5747\u7372\u5f97\u986f\u8457\u7684\u63d0\u5347\u3002 \u672c\u8ad6\u6587\u63d0\u51fa\u5169\u500b\u65b0\u7a4e\u7684\u91cd\u5beb\u5f0f\u6458\u8981\u6a21\u578b\u3002\u7b2c\u4e00\u500b\u6a21\u578b\u4ee5\u6307\u91dd\u751f\u6210\u7db2\u8def\u70ba\u57fa\u790e\uff0c\u52a0\u4e0a BERT \u4f5c\u70ba\u7de8\u78bc\u5668\uff0c\u671f\u671b BERT \u80fd\u751f\u6210\u5f37\u5065\u4e14\u6e96\u78ba\u7684\u6587\u7ae0\u8868\u793a\u7279\u5fb5\uff0c\u63d0\u5347\u91cd\u5beb\u5f0f\u6458\u8981\u7684 \u4efb\u52d9\u6210\u6548\uff0c\u6211\u5011\u7a31\u70ba\u300c\u4ee5 BERT \u70ba\u7de8\u78bc\u5668\u4e4b\u6307\u91dd\u751f\u6210\u6458\u8981\u6cd5\u300d \u3002\u7b2c\u4e8c\u500b\u6a21\u578b\u70ba\u7b2c\u4e00\u500b\u6a21 \u578b\u7684\u5ef6\u4f38\uff0c\u9664\u4e86\u4f7f\u7528 BERT \u4f5c\u70ba\u7de8\u78bc\u5668\u5916\uff0c\u6211\u5011\u4f7f\u7528 Transformer \u67b6\u69cb\u4ee3\u66ff\u50b3\u7d71\u905e\u8ff4\u795e", |
| "type_str": "table", |
| "content": "<table><tr><td colspan=\"6\">\u5668\u7684\u91cd\u8981\u6027\u3002\u5728\u89e3\u78bc\u5668\u4e2d\uff0c\u905e\u8ff4\u795e\u7d93\u7db2\u8def\u6703\u5728\u6bcf\u500b\u6642\u9593\u9ede\u7522\u751f\u4e00\u500b\u8f38\u51fa\u5411\u91cf \uff0c\u6b64\u4e00\u5411</td></tr><tr><td colspan=\"6\">\u91cf \u5c07 \u8207 \u7de8 \u78bc \u5668 \u7684\u6bcf \u500b\u6642 \u9593 \u9ede \u8f38 \u51fa \u210e \u8a08 \u7b97 \u5f97 \u5230\u4e00 \u500b \u76f8 \u95dc \u6027 \u6b0a \u91cd \uff0c \u4e26\u900f \u904e \u6b63 \u898f \u5316</td></tr><tr><td colspan=\"6\">(Normalize)\uff0c\u7372\u5f97 \u6642\u9593\u9ede\uff0c\u89e3\u78bc\u5668\u5c0d\u65bc\u7de8\u78bc\u5668\u7684\u6ce8\u610f\u529b\u6b0a\u91cd =</td><td>( )\uff0c\u5176\u4e2d</td></tr><tr><td>, \u210e , ,</td><td colspan=\"5\">\u5373\u70ba\u6ce8\u610f\u529b\u6a5f\u5236\u4e2d\u7684\u6a21\u578b\u53c3\u6578\u3002\u63a5\u8457\uff0c\u5c07\u6bcf\u4e00\u500b\u6ce8\u610f\u529b\u6b0a\u91cd \u8207\u6240\u5c0d\u61c9</td></tr><tr><td colspan=\"6\">\u7684\u210e \u76f8\u4e58\u5f8c\u52a0\u7e3d\uff0c\u5c31\u53ef\u7372\u5f97\u7576\u524d\u6642\u9593\u9ede\u4e4b\u6ce8\u610f\u529b\u5411\u91cf * :</td></tr><tr><td/><td/><td>=</td><td/><td>\u210e ( \u210e \u210e +</td><td>+</td><td>)</td><td>\u5f0f(1)</td></tr><tr><td/><td/><td>* = \u2211</td><td>=1</td><td>\u210e</td><td>\u5f0f(2)</td></tr><tr><td colspan=\"2\">\u518d\u7d93\u904e\u5168\u9023\u63a5\u5c64\u4ee5\u53ca</td><td colspan=\"4\">\u6fc0\u6d3b\u51fd\u6578\uff0c\u6703\u8f38\u51fa\u4e00\u500b\u8fad\u5178\u5927\u5c0f\u7dad\u5ea6\u7684\u6a5f\u7387\u5206\u5e03\uff0c\u6bcf\u4e00\u500b</td></tr><tr><td colspan=\"6\">\u7dad\u5ea6\u5c0d\u61c9\u4e00\u500b\u5b57\u5178\u4e2d\u7684\u8a5e\u3002\u6211\u5011\u5c07\u5206\u6578\u6700\u9ad8\u7684\u8a5e\u9078\u53d6\u51fa\u4f86\uff0c\u4f5c\u70ba\u6b64\u6642\u9593\u9ede\u7684\u8f38\u51fa\uff0c\u540c\u6642</td></tr><tr><td colspan=\"6\">\u4e5f\u662f\u4e0b\u500b\u6642\u9593\u9ede\u7684\u8f38\u5165\u3002\u91cd\u8907\u6b65\u9a5f\u76f4\u5230\u89e3\u78bc\u5668\u8f38\u51fa\u7279\u6b8a\u7b26\u865f < \u900f\u904e\u7dda\u6027\u7d44\u5408\uff0c\u7522\u751f\u89e3\u78bc\u5668\u5728\u6642\u9593\u9ede \u7684\u53c3\u8003\u6a5f\u7387\u5206\u5e03 ( )\u3002\u85c9\u7531\u6b64\u65b9\u5f0f\uff0c\u4e0d\u540c\u65bc\u5e8f\u5217 >\u70ba\u6b62\u3002</td></tr><tr><td colspan=\"6\">(\u4e8c)\u6307\u91dd\u751f\u6210\u7db2\u8def(Pointer Generator Network, PGN) \u5c0d\u5e8f\u5217\u6a21\u578b\uff0c\u6307\u91dd\u751f\u6210\u7db2\u8def\u4fbf\u53ef\u4ee5\u8f38\u51fa\u975e\u5b57\u5178\u88e1\u7684\u8a5e\u3002</td></tr><tr><td colspan=\"6\">\u7d93\u7db2\u8def\uff0c\u8b93\u6ce8\u610f\u529b\u6a5f\u5236\u4f86\u7372\u5f97\u6642\u9593\u5e8f\u5217\u4e0a\u7684\u95dc\u4fc2\uff0c\u7528\u4ee5\u89e3\u6c7a\u9577\u6642\u9593\u5e8f\u5217\u8a0a\u606f\u4e1f\u5931\u554f\u984c\uff0c \u5728\u50b3\u7d71\u5e8f\u5217\u5c0d\u5e8f\u5217\u4e4b\u91cd\u5beb\u5f0f\u6458\u8981\u7684\u4efb\u52d9\u4e2d\uff0c\u67d0\u4e9b\u8a5e\u5f59\u96d6\u7136\u51fa\u73fe\u5728\u8f38\u5165\u6587\u5b57\u5e8f\u5217\u4e2d\uff0c\u4f46\u82e5 ( ) = \u2211 : = \u5f0f(4)</td></tr><tr><td colspan=\"6\">\u7522\u751f\u66f4\u52a0\u5f37\u5065\u4e14\u4f9d\u8cf4\u4e0a\u4e0b\u6587\u8cc7\u8a0a\u7684\u7279\u5fb5\uff0c\u671f\u671b\u8b93\u91cd\u5beb\u5f0f\u6458\u8981\u6548\u80fd\u518d\u9032\u4e00\u6b65\u63d0\u5347\uff0c\u6211\u5011\u7a31 \u5b83\u4e26\u975e\u5b58\u5728\u5b57\u5178\u4e2d(Out-of-Vocabulary, OOV)\uff0c\u9019\u985e\u8a5e\u5f59\u662f\u7121\u6cd5\u88ab\u89e3\u78bc\u5668\u89e3\u78bc\u51fa\u4f86\u7684\u3002\u70ba</td></tr><tr><td colspan=\"6\">\u70ba\u300c\u878d\u5408 BERT \u8207 Transformer \u4e4b\u6307\u91dd\u751f\u6210\u6458\u8981\u6cd5\u300d \u3002\u6b64\u5916\uff0c\u6211\u5011\u5c07\u63a2\u8a0e\u9019\u4e9b\u6a21\u578b\u5728\u4e2d\u6587 \u4e86\u89e3\u6c7a\u6b64\u4e00\u554f\u984c\uff0c\u8fd1\u5e74\u6709\u7814\u7a76\u63d0\u51fa\u4e86\u6307\u91dd\u751f\u6210\u7db2\u8def(Pointer Generator Network, PGN)\uff0c\u5176</td></tr><tr><td colspan=\"6\">\u91cd\u5beb\u5f0f\u6458\u8981\u7684\u4efb\u52d9\u6210\u6548\uff0c\u4ee5\u4f5c\u70ba\u5f8c\u7e8c\u7814\u7a76\u7684\u91cd\u8981\u6bd4\u8f03\u57fa\u790e\u3002 \u67b6\u69cb\u5982\u5716\u4e00\u6240\u793a\u3002\u4e00\u7bc7\u6587\u7ae0\u4e4b\u8a5e\u5f59{ 1 , 2 \u2026 , , \u2026 }\uff0c\u9010\u4e00\u8f38\u5165\u7de8\u78bc\u5668\u905e\u8ff4\u795e\u7d93\u7db2\u8def</td></tr><tr><td colspan=\"6\">\u5f8c\uff0c\u7522\u751f\u6bcf\u500b\u6642\u9593\u9ede\u7684\u8f38\u51fa\u5411\u91cf\u210e \u3002\u8207\u5e8f\u5217\u5c0d\u5e8f\u5217\u6a21\u578b\u4e2d\u7684\u89e3\u78bc\u5668\u4e00\u6a23\uff0c\u905e\u8ff4\u795e\u7d93\u7db2\u8def \u4e8c\u3001\u76f8\u95dc\u65b9\u6cd5 \u6703\u5728\u6bcf\u500b\u6642\u9593\u9ede\u7522\u751f\u8f38\u51fa \uff0c\u4e26\u8207\u7de8\u78bc\u5668\u6bcf\u500b\u6642\u9593\u9ede\u210e \u7522\u751f \uff0c\u85c9\u7531\u6b63\u898f\u5316\u5f97\u5230\u6ce8\u610f\u529b</td></tr><tr><td colspan=\"6\">\u6b0a\u91cd \uff0c\u7136\u5f8c\u6211\u5011\u5c07\u6ce8\u610f\u529b\u6b0a\u91cd = [ 1 , \u2026 , , \u2026 , ]\u4e58\u4e0a\u7de8\u78bc\u5668\u6bcf\u500b\u6642\u9593\u9ede\u4e4b\u210e \uff0c\u76f8</td></tr><tr><td colspan=\"6\">(\u4e00)\u5e8f\u5217\u5c0d\u5e8f\u5217\u6a21\u578b \u52a0\u5f8c\u5f97\u5230\u7576\u524d\u6642\u9593\u9ede\u89e3\u78bc\u5668\u4e4b\u6ce8\u610f\u529b\u5411\u91cf * \uff0c\u518d\u900f\u904e\u5168\u9023\u63a5\u5c64\u4ee5\u53ca</td><td>\u6fc0\u6d3b\u51fd\u6578\uff0c</td></tr><tr><td colspan=\"6\">\u5e8f\u5217\u5c0d\u5e8f\u5217\u6a21\u578b(Sequence-to-sequence)[7]\u4e3b\u8981\u5305\u542b\u5169\u5927\u90e8\u5206\uff0c\u7de8\u78bc\u5668(Encoder)\u8207\u89e3\u78bc\u5668 \u53ef\u4ee5\u5f97\u5230\u5b57\u5178(Vocabulary)\u4e2d\u6240\u6709\u8a5e\u5f59\u7684\u6a5f\u7387\u5206\u5e03 \uff1a</td></tr><tr><td colspan=\"6\">(Decoder)\u3002\u7de8\u78bc\u5668\u8207\u89e3\u78bc\u5668\u5927\u591a\u7531\u905e\u8ff4\u795e\u7d93\u7db2\u8def(Recurrent Neural Networks)\u69cb\u6210\uff0c\u4f8b\u5982 \u9577\u77ed\u671f\u8a18\u61b6\u7db2\u8def(Long Short-term Memory, LSTM)[8]\u3002 = (\u0300( [ , * ] + ) +\u0300) \u5f0f(3)</td></tr><tr><td colspan=\"6\">\u5176\u4e2d\uff0c\u0300, , ,\u0300\u70ba\u53ef\u5b78\u7fd2\u4e4b\u6a21\u578b\u53c3\u6578\u3002\u9664\u4e86 \u7d66\u5b9a\u4e00\u6bb5\u8a5e\u5f59\u5e8f\u5217{ 1 , 2 \u2026 , , \u2026 }\u4f9d\u5e8f\u8f38\u5165\u7de8\u78bc\u5668\u4e2d\uff0c\u6bcf\u500b\u8a5e\u5f59 \u4e4b\u8a5e\u5411\u91cf \u5916\uff0c\u6307\u91dd\u751f\u6210\u7db2\u8def\u6703\u7522\u751f\u4e00\u500b\u50c5\u7531\u8f38\u5165 \u7684\u6587\u5b57\u5e8f\u5217\u8a08\u7b97\u800c\u5f97\u7684\u8a9e\u8a00\u6a21\u578b \uff0c\u9019\u500b\u8a9e\u8a00\u6a21\u578b\u7684\u8fad\u5178\u50c5\u7531\u8f38\u5165\u4e2d\u6240\u6709\u4e0d\u540c\u7684\u5b57\u8a5e \u6703\u8207\u524d\u4e00\u500b\u6642\u9593\u9ede\u905e\u8ff4\u795e\u7d93\u7db2\u8def\u7684\u8f38\u51fa\u210e \u22121 \uff0c\u4e00\u8d77\u8f38\u5165\u905e\u8ff4\u795e\u7d93\u7db2\u8def\uff0c\u7522\u751f\u6b64\u6642\u9593\u9ede\u905e \u8ff4\u795e\u7d93\u7db2\u8def\u7684\u8f38\u51fa\u210e \u3002\u5728\u89e3\u78bc\u5668\u90e8\u5206\uff0c\u7531\u65bc\u8f38\u5165\u6587\u7ae0\u7684\u6bcf\u500b\u8a5e\u5f59\u5c0d\u65bc\u89e3\u78bc\u5668\u7522\u751f\u7684\u6bcf\u500b \u6240\u7d44\u6210\uff0c\u56e0\u6b64\u53ef\u80fd\u5305\u542b\u6c92\u6709\u51fa\u73fe\u5728 \u4e2d\u7684\u5b57\u8a5e\u3002 \u53ef\u4ee5\u5feb\u901f\u5730\u7531\u6ce8\u610f\u529b\u6b0a\u91cd \u8a08</td></tr><tr><td colspan=\"6\">\u8f38\u51fa\u91cd\u8981\u7a0b\u5ea6\u4e26\u4e0d\u4e00\u6a23\uff0c\u6709\u7814\u7a76\u63d0\u51fa\u52a0\u5165\u6ce8\u610f\u529b\u6a5f\u5236[9]\uff0c\u4f7f\u5f97\u89e3\u78bc\u5668\u5728\u6bcf\u500b\u6642\u9593\u9ede\uff0c\u6703 \u7b97\u800c\u5f97\uff0c\u7531\u65bc\u6ce8\u610f\u529b\u6b0a\u91cd \u5df2\u7d93\u904e\u6b63\u898f\u5316\uff0c\u56e0\u6b64 \u5fc5\u5b9a\u6eff\u8db3\u6a5f\u7387\u516c\u8a2d\u2211 ( ) = 1\u3002</td></tr><tr><td colspan=\"6\">\u5c0d\u7de8\u78bc\u5668\u7684\u6240\u6709\u6642\u9593\u9ede\u7522\u751f\u4e00\u500b\u6ce8\u610f\u529b\u6b0a\u91cd\uff0c\u8868\u793a\u7de8\u78bc\u5668\u4e2d\u6bcf\u4e00\u500b\u6642\u9593\u9ede\u5c0d\u65bc\u6b64\u6642\u89e3\u78bc \u6700\u5f8c\uff0c\u6211\u5011\u5229\u7528 \u3001 * \u548c\u89e3\u78bc\u5668\u6642\u9593\u9ede \u4e4b\u8f38\u5165 \u8a08\u7b97\u51fa \u8207 \u7684\u7d50\u5408\u4fc2\u6578 \uff0c\u4e26</td></tr></table>" |
| } |
| } |
| } |
| } |
