{
    "paper_id": "2020",
    "header": {
        "generated_with": "S2ORC 1.0.0",
        "date_generated": "2023-01-19T12:53:44.203470Z"
    },
    "title": "Research on Response Generation via Dialogue Constraints",
    "authors": [
        {
            "first": "Mengyu",
            "middle": [],
            "last": "Guan\uff0c",
            "suffix": "",
            "affiliation": {
                "laboratory": "",
                "institution": "Soochow University",
                "location": {
                    "settlement": "Suzhou",
                    "country": "China"
                }
            },
            "email": ""
        },
        {
            "first": "Wang\uff0c",
            "middle": [
                "\uff0c"
            ],
            "last": "\uff0cshoushan",
            "suffix": "",
            "affiliation": {
                "laboratory": "",
                "institution": "Soochow University",
                "location": {
                    "settlement": "Suzhou",
                    "country": "China"
                }
            },
            "email": ""
        },
        {
            "first": "Li\uff0c",
            "middle": [
                "\uff0c"
            ],
            "last": "\uff0cguodong Zhou",
            "suffix": "",
            "affiliation": {
                "laboratory": "",
                "institution": "Soochow University",
                "location": {
                    "settlement": "Suzhou",
                    "country": "China"
                }
            },
            "email": ""
        }
    ],
    "year": "",
    "venue": null,
    "identifiers": {},
    "abstract": "Existing dialogue systems tend to generate meaningless general replies such as \"OK\" and \"I don't know\". In daily dialogs, every utterance usually has obvious emotional and intentional tendencies. So this paper proposes a response generation model based on dialogue constraints. Based on the Seq2Seq model, it combines the recognition of utterances' themes, sentiments and intentions. This method constrains the topics, emotions and intentions of the generated responses, generating responses with reasonable sentiment and intention tendencies and related to the topic of the conversation. Experiments show that the method proposed in this paper can effectively improve the quality of generated responses.",
    "pdf_parse": {
        "paper_id": "2020",
        "_pdf_hash": "",
        "abstract": [
            {
                "text": "Existing dialogue systems tend to generate meaningless general replies such as \"OK\" and \"I don't know\". In daily dialogs, every utterance usually has obvious emotional and intentional tendencies. So this paper proposes a response generation model based on dialogue constraints. Based on the Seq2Seq model, it combines the recognition of utterances' themes, sentiments and intentions. This method constrains the topics, emotions and intentions of the generated responses, generating responses with reasonable sentiment and intention tendencies and related to the topic of the conversation. Experiments show that the method proposed in this paper can effectively improve the quality of generated responses.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Abstract",
                "sec_num": null
            }
        ],
        "body_text": [
            {
                "text": "\u4eba\u673a\u4ea4\u4e92(Human Computer Interaction, HCI)\u4f5c\u4e3a\u4fe1\u606f\u65f6\u4ee3\u4eba\u7c7b\u4e0e\u8ba1\u7b97\u673a\u4e4b\u95f4\u4fe1\u606f\u4ea4\u6d41 \u7684\u57fa\u7840\u6280\u672f\uff0c\u53d7\u5230\u5b66\u672f\u754c\u548c\u5de5\u4e1a\u754c\u7684\u5e7f\u6cdb\u5173\u6ce8\u3002\u4eba\u673a\u5bf9\u8bdd(Human-Machine Dialogue)\u662f\u4eba\u673a \u4ea4\u4e92\u6280\u672f\u7684\u6838\u5fc3\u9886\u57df\uff0c\u65e8\u5728\u6700\u5927\u9650\u5ea6\u5730\u6a21\u4eff\u4eba\u4e0e\u4eba\u4e4b\u95f4\u7684\u5bf9\u8bdd\u65b9\u5f0f\uff0c\u4f7f\u5f97\u4eba\u7c7b\u80fd\u591f\u7528\u66f4\u81ea \u7136\u7684\u65b9\u5f0f\u4e0e\u673a\u5668\u8fdb\u884c\u4ea4\u6d41\u3002\u5176\u5e94\u7528\u573a\u666f\u5e7f\u6cdb\uff0c\u5177\u6709\u8f83\u9ad8\u7684\u7814\u7a76\u4ef7\u503c\u548c\u5546\u4e1a\u4ef7\u503c\u3002\u6784\u5efa\u4e00\u4e2a \u8f83\u5b8c\u5907\u7684\u4eba\u673a\u5bf9\u8bdd\u7cfb\u7edf\u6d89\u53ca\u5230NLP\u6280\u672f\u7684\u5f88\u591a\u65b9\u9762\uff0c\u6bd4\u5982\u53e5\u6cd5\u5206\u6790 (Fried et al., 2017 )\uff0c\u547d \u540d\u5b9e\u4f53\u8bc6\u522b (Huang et al., 2015) \u7b49\u3002\u672c\u6587\u4e3b\u8981\u7814\u7a76\u591a\u8f6e\u5bf9\u8bdd\u7684\u56de\u590d\u751f\u6210\u3002\u7b80\u800c\u8a00\u4e4b\u5c31\u662f\u6839\u636e \u5386\u53f2\u5bf9\u8bdd\u4fe1\u606f\uff0c\u81ea\u52a8\u751f\u6210\u81ea\u7136\u5408\u7406\u7684\u56de\u590d\uff0c\u5728\u4fe1\u606f\u4ea4\u4e92\u7684\u8fc7\u7a0b\u4e2d\u534f\u52a9\u7528\u6237\u5b8c\u6210\u7279\u5b9a\u7684\u4efb\u52a1\u3002 \u968f\u7740\u7aef\u5230\u7aef\u6846\u67b6\u5728\u673a\u5668\u7ffb\u8bd1 (Brown et al., 1993) \u4efb\u52a1\u4e0a\u7684\u826f\u597d\u8868\u73b0\uff0c\u7814\u7a76\u4eba\u5458\u5c06\u5176\u8fc1\u79fb\u5e94\u7528 \u4e8e\u5bf9\u8bdd\u751f\u6210\u4efb\u52a1\u4e2d\u3002\u5bf9\u8bdd\u751f\u6210\u53ef\u4ee5\u7b80\u5316\u4e3a\u8f93\u5165\u8f93\u51fa\u7684\u6620\u5c04\u95ee\u9898\uff0c\u5373\u5bf9\u5bf9\u8bdd\u7684\u8f93\u5165\u8fdb\u884c\u7f16\u7801\u548c \u89e3\u7801\u4ece\u800c\u5f97\u5230\u5e94\u7b54\u3002\u56e0\u4e3a\u5bf9\u8bdd\u662f\u6709\u65f6\u5e8f\u7684\uff0c\u53ef\u4ee5\u89c6\u4e3a\u5e8f\u5217\uff0c\u6240\u4ee5\u7aef\u5bf9\u7aef\u6846\u67b6\u4e0b\u7684\u5e8f\u5217\u5bf9\u5e8f\u5217 \u6a21(Sequence-to-Sequence, Seq2Seq) (Sutskever et al., 2014; Cho et al., 2014) \u975e\u5e38\u9002\u5408\u5bf9\u8bdd\u751f\u6210 \u6a21\u578b\u3002 \u4e3b\u9898\u7b80\u8981\u5730\u8868\u660e\u4e86\u6574\u6bb5\u5bf9\u8bdd\u7684\u5185\u5bb9\uff0c\u7528\u4e8e\u7406\u89e3\u5bf9\u8bdd\u7684\u542b\u4e49\uff1b\u60c5\u611f\u8868\u660e\u4e86\u8bdd\u8bed\u7684\u6781\u6027\uff0c\u7528 \u4e8e\u8bc6\u522b\u8bf4\u8bdd\u8005\u7684\u89c2\u70b9\uff1b\u610f\u56fe\u4f5c\u4e3a\u8bdd\u8bed\u7684\u8bed\u4e49\u6807\u7b7e\uff0c\u7528\u4e8e\u63cf\u8ff0\u8bf4\u8bdd\u8005\u7684\u52a8\u4f5c\u610f\u56fe\u3002\u539f\u5219\u4e0a\uff0c\u5bf9 \u4e3b\u9898\u3001\u60c5\u611f\u548c\u610f\u56fe\u7684\u8bc6\u522b\u6709\u52a9\u4e8e\u5bf9\u8bdd\u8bed\u5185\u5bb9\u7684\u7406\u89e3\u3002\u8868 1\u7ed9\u51fa\u4e86\u4e00\u7ec4\u5bf9\u8bdd\u793a\u4f8b\uff0c\u6574\u6bb5\u5bf9\u8bdd\u662f\u5173 \u4e8e\"\u65e5\u5e38\u751f\u6d3b\"\u7684\u8ba8\u8bba\u4e14\u5bf9\u8bdd\u4e2d\u7684\u6bcf\u53e5\u8bdd\u90fd\u6709\u660e\u786e\u7684\u60c5\u611f\u548c\u610f\u56fe\u503e\u5411\u3002\u6211\u4eec\u53d1\u73b0\uff0c\u540c\u4e00\u8bf4\u8bdd\u8005\u7684 \u60c5\u611f\u5f80\u5f80\u662f\u4fdd\u6301\u4e0d\u53d8\u7684\uff1b\u8bf4\u8bdd\u8005B\u4f5c\u4e3a\u5bf9\u8bdd\u4e2d\u88ab\u52a8\u7684\u4e00\u65b9\uff0c\u610f\u56fe\u5f80\u5f80\u53d7\u5230\u5bf9\u8bdd\u53d1\u8d77\u8005A\u7684\u610f\u56fe \u7684\u5f71\u54cd\u3002\u56e0\u6b64\u672c\u6587\u901a\u8fc7\u8bc6\u522b\u5bf9\u8bdd\u7684\u4e3b\u9898\u53ca\u6bcf\u4e2a\u8bdd\u8bed\u7684\u60c5\u611f\u548c\u610f\u56fe\uff0c\u5bf9\u751f\u6210\u7684\u56de\u590d\u8fdb\u884c\u7ea6\u675f\u3002 \u82e5\u6211\u4eec\u8981\u751f\u6210\u7684\u662f\u7b2c\u4e8c\u8f6e\u4e2d\u8bf4\u8bdd\u8005B\u7684\u56de\u590d\uff0c\u6211\u4eec\u7684\u65b9\u6cd5\u5c06\u751f\u6210\u4e0e\u5bf9\u8bdd\u4e3b\u9898\u76f8\u5173\u3001\u60c5\u611f\u503e\u5411 \u4e8e\"\u4e2d\u6027\"\u3001\u610f\u56fe\u503e\u5411\u4e8e\"\u627f\u8bfa\"\u7684\u56de\u590d\uff0c\u800c\u4e0d\u662f\u6211\u4eec\u901a\u5e38\u5f97\u5230\u7684\u4f8b\u5982\"\u597d\u7684\"\uff0c\"\u6211\u4e0d\u77e5\u9053\"\u7b49\u5b89\u5168 \u56de\u590d (McKay and Piperno, 2014) ",
                "cite_spans": [
                    {
                        "start": 220,
                        "end": 239,
                        "text": "(Fried et al., 2017",
                        "ref_id": "BIBREF4"
                    },
                    {
                        "start": 250,
                        "end": 270,
                        "text": "(Huang et al., 2015)",
                        "ref_id": "BIBREF7"
                    },
                    {
                        "start": 352,
                        "end": 372,
                        "text": "(Brown et al., 1993)",
                        "ref_id": "BIBREF0"
                    },
                    {
                        "start": 508,
                        "end": 532,
                        "text": "(Sutskever et al., 2014;",
                        "ref_id": "BIBREF20"
                    },
                    {
                        "start": 533,
                        "end": 550,
                        "text": "Cho et al., 2014)",
                        "ref_id": "BIBREF1"
                    },
                    {
                        "start": 903,
                        "end": 928,
                        "text": "(McKay and Piperno, 2014)",
                        "ref_id": "BIBREF13"
                    }
                ],
                "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": "3 \u57fa \u57fa \u57fa\u4e8e \u4e8e \u4e8e\u5bf9 \u5bf9 \u5bf9\u8bdd \u8bdd \u8bdd\u7ea6 \u7ea6 \u7ea6\u675f \u675f \u675f\u7684 \u7684 \u7684\u56de \u56de \u56de\u590d \u590d \u590d\u751f \u751f \u751f\u6210 \u6210 \u6210\u6a21 \u6a21 \u6a21\u578b \u578b \u578b \u6211\u4eec\u7684\u4efb\u52a1\u65e8\u5728\u81ea\u52a8\u751f\u6210\u5408\u7406\u81ea\u7136\u7684\u5bf9\u8bdd\u56de\u590d\u3002\u4e00\u7ec4\u5bf9\u8bdd\u7531\u4e24\u4e2a\u5bf9\u8bdd\u8005\u4e4b\u95f4\u53d1\u8d77\u7684m/2\u8f6e \u5bf9\u8bdd\u7ec4\u6210\uff0c\u53ef\u8868\u793a\u4e3a\u5bf9\u8bdd\u5e8f\u5217D = {U 1 , U 2 , ..., U m }\uff0c\u5176\u4e2dU x (x = 1, 2. . . )\u79f0\u4e3a\u5bf9\u8bdd\u7684\u5b50\u53e5\u3002\u5bf9 \u8bdd\u751f\u6210\u6a21\u578b\u7684\u76ee\u7684\u662f\u5728\u7b2cm/2\u8f6e\u65f6\uff0c\u6839\u636e\u524d\u9762\u7684m \u2212 1\u4e2a\u5b50\u53e5{U 1 , U 2 , ..., U m\u22121 }\u8ba1\u7b97\u5728\u6b64\u60c5\u51b5 \u4e0b\u751f\u6210\u53e5\u5b50U m \u7684\u6982\u7387\uff0c\u5373P (U m |U 1 , U 2 , ..., U m\u22121 )\u3002\u6bcf\u4e2a\u53e5\u5b50U m \u662f\u53ef\u53d8\u957f\u7684\u5355\u8bcd\u5e8f\u5217\uff0c\u53ef\u8868\u793a \u4e3aU m = {w m,1 , w m,2 , ..., w m,Nm }\u3002\u5176\u4e2dw m,n \u8868\u793a\u7b2cm\u4e2a\u53e5\u5b50\u4e2d\u7684\u7b2cn\u4e2a\u5355\u8bcd\uff0cN m \u8868\u793aU m \u4e2d\u7684\u5355 \u8bcd\u4e2a\u6570\u3002\u901a\u8fc7\u524dm-1\u4e2a\u5b50\u53e5\u548c\u5f53\u524d\u5df2\u7ecf\u751f\u6210\u7684\u5355\u8bcd\u6765\u9010\u5b57\u9884\u6d4b\u4e0b\u4e00\u4e2a\u8bcd\uff0c\u76f4\u5230\u8fbe\u5230\u7279\u5b9a\u7684\u53e5\u5b50 \u957f\u5ea6\u6216\u8005\u751f\u6210\u7ed3\u675f\u7b26\uff0c\u9884\u6d4b\u7ed3\u675f\uff0c\u5f97\u5230\u56de\u590dU m \u3002P (U m |U 1 , U 2 , ..., U m\u22121 )\u53ef\u8868\u793a\u4e3a\u5982\u4e0b\u516c\u5f0f\uff1a P (w m,1 , w m,2 , ...., w m,n |U 1 , U 2 , ..., U m\u22121 ) = Nm n=1 P (w m,n |U 1 , U 2 , .., U m\u22121 , w <n )",
                        "eq_num": "(1)"
                    }
                ],
                "section": "",
                "sec_num": null
            },
            {
                "text": "3.1 \u6a21 \u6a21 \u6a21\u578b \u578b \u578b \u672c\u6587\u91c7\u7528\u7684\u662f\u57fa\u4e8eLSTM\u7f51\u7edc\u7684Seq2Seq\u6a21\u578b\u3002\u8be5\u6a21\u578b\u5305\u62ec3\u4e2a\u90e8\u5206: (Papineni et al., 2002 )\u6765\u8861\u91cf\u6240\u4ea7\u751f\u7684 \u54cd\u5e94\u4e0e\u5730\u9762\u771f\u5b9e\u54cd\u5e94\u4e4b\u95f4\u7684\u76f8\u5173\u6027\u3002\u672c\u6587\u4e3b\u8981\u4f7f\u7528BLEU-1\u3001BLEU-2\u548cBLEU-3\u6765\u8bc4\u6d4b\u5b9e\u9a8c\u6548 \u679c\u3002\u672c\u6587\u8fd8\u4f7f\u7528\u8bcd\u5d4c\u5165\u4e4b\u95f4\u7684\u4f59\u5f26\u76f8\u4f3c\u5ea6\u6765\u5ea6\u91cf\u8bcd\u4e4b\u95f4\u7684\u76f8\u4f3c\u6027\uff0c\u53ef\u5206\u4e3a\u4e09\u7c7b\uff1aAverage (Foltz et al., 1998) \uff0cGreedy (Rus and Lintean, 2012) \u548cExtrema (Forgues et al., 2014) ",
                "cite_spans": [
                    {
                        "start": 51,
                        "end": 73,
                        "text": "(Papineni et al., 2002",
                        "ref_id": "BIBREF15"
                    },
                    {
                        "start": 175,
                        "end": 195,
                        "text": "(Foltz et al., 1998)",
                        "ref_id": "BIBREF2"
                    },
                    {
                        "start": 204,
                        "end": 227,
                        "text": "(Rus and Lintean, 2012)",
                        "ref_id": "BIBREF16"
                    },
                    {
                        "start": 237,
                        "end": 259,
                        "text": "(Forgues et al., 2014)",
                        "ref_id": "BIBREF3"
                    }
                ],
                "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": "\u7f16\u7801\u5668( Encoder) \u7f51 \u7edc\u3001\u89e3\u7801\u5668( Decoder)\u7f51\u7edc\u548c\u8fde\u63a5Encoder-Decoder\u7684\u4e2d\u95f4\u8bed\u4e49\u5411\u91cfC\u3002\u7f16\u7801\u5668\u7f51\u7edc\u548c\u89e3\u7801\u5668\u7f51 \u7edc\u5206\u522b\u5bf9\u5e94\u8f93\u5165\u5e8f\u5217\u548c\u8f93\u51fa\u5e8f\u5217\u7684\u4e24\u4e2a\u795e\u7ecf\u7f51\u7edc\uff0c\u672c\u5b9e\u9a8c\u4e2d\u7f16\u7801\u5668\u548c\u89e3\u7801\u5668\u662f\u57fa\u4e8eLSTM\u7f51\u7edc \u5904\u7406\u8f93\u5165\u5e8f\u5217\u548c\u8f93\u51fa\u5e8f\u5217\u7684\u3002\u8f93\u5165\u5e8f\u5217\u901a\u8fc7\u7f16\u7801\u5668\u5bf9\u5176\u8fdb\u884c\u7f16\u7801\uff0c\u5f62\u6210\u4e00\u4e2a\u4e2d\u95f4\u8bed\u4e49\u5411\u91cfC\uff0c LSTM ... LSTM LSTM LSTM 1 2 3 \u22121 1 2 3 \u22121 ... \u4e2d\u95f4\u8bed\u4e49\u5411\u91cf LSTM \u7f16\u7801\u5668 \u89e3\u7801\u5668 Topic Sentiment Act Topic of dialog D Sentiment of 1 , 2 , \u2026 Act of 1 , 2 , \u2026 \u56de\u590d\u751f\u6210\u6a21\u578b \u60c5\u611f\u3001\u8bdd\u9898\u3001\u610f\u56fe\u9884\u6d4b\u6a21\u578b \u56fe 1. \u6a21\u578b\u7f51\u7edc\u7ed3\u6784\u56fe \u628a\u8fd9\u4e2a\u5411\u91cf\u4f20\u9012\u7ed9\u89e3\u7801\u5668\uff0c\u901a\u8fc7\u4e2d\u95f4\u8bed\u4e49\u5411\u91cfC\u548c\u9012\u5f52\u9690\u85cf\u72b6\u6001\uff0c\u8ba1\u7b97\u8bcd\u6c47\u8868\u4e2d\u6bcf\u4e2a\u8bcd\u7684\u6982\u7387 \u5206\u5e03\uff0c\u751f\u6210\u4e0b\u4e00\u4e2a\u8bcd\uff0c\u5e76\u5c06\u5176\u4f5c\u4e3a\u4e0b\u4e00\u4e2a\u65f6\u523b\u7684\u8f93\u5165\uff0c\u76f4\u81f3\u9047\u5230\u7ed3\u675f\u7b26\u6807\u5fd7\u6216\u8fbe\u5230\u6307\u5b9a\u7684\u957f\u5ea6 \u65f6\u7ed3\u675f\u89e3\u7801\u3002\u53e6\u5916\u6211\u4eec\u7684\u6a21\u578b(JTEA)\u5728\u6a21\u578b\u8bad\u7ec3\u9636\u6bb5\uff0c\u4ece\u7f16\u7801\u5668\u548c\u89e3\u7801\u5668\u7684LSTM\u7f51\u7edc\u4e2d\u5b66\u4e60 \u5bf9\u8bdd\u5386\u53f2\u5bf9\u8bdd\u4fe1\u606f\u4e2d\u6bcf\u4e2a\u5b50\u53e5\u548c\u751f\u6210\u56de\u590d\u7684\u7279\u5f81\u8868\u793a\u3002\u8fd9\u4e9b\u7279\u5f81\u5411\u91cf\u7528\u4e8e\u9884\u6d4b\u6bcf\u4e2a\u8bdd\u8bed\u7684\u60c5\u611f \u548c\u610f\u56fe\u4ee5\u53ca\u6574\u6bb5\u5bf9\u8bdd\u7684\u4e3b\u9898\u3002\u4e3b\u9898\u9884\u6d4b\u3001\u60c5\u611f\u9884\u6d4b\u548c\u610f\u56fe\u9884\u6d4b\u4f5c\u4e3a\u5bf9\u8bdd\u56de\u590d\u751f\u6210\u7684\u8f85\u52a9\u4efb\u52a1\uff0c \u53ea\u5728\u6a21\u578b\u8bad\u7ec3\u9636\u6bb5\u8fdb\u884c\uff0c\u5e2e\u52a9\u5b66\u4e60\u7f16\u7801\u5668\u548c\u89e3\u7801\u5668\u4e2d\u7684\u5171\u4eab\u53c2\u6570\uff0c\u5bf9\u751f\u6210\u56de\u590d\u7684\u4e3b\u9898\u3001\u60c5\u611f\u548c \u610f\u56fe\u8fdb\u884c\u7ea6\u675f\u3002\u6a21\u578b\u7f51\u7edc\u7ed3\u6784\u56fe\u5982\u56fe 1\u6240\u793a\u3002 3.1.1 \u7f16 \u7f16 \u7f16\u7801 \u7801 \u7801\u5668 \u5668 \u5668 \u7f16\u7801\u5668\u662f\u5c06\u8f93\u5165\u5e8f\u5217\u7f16\u7801\u6210\u4e00\u4e2a\u4e2d\u95f4\u8bed\u4e49\u5411\u91cfC\u3002\u672c\u5b9e\u9a8c\u4e2d\u4f7f\u6bcf\u4e2a\u5b50\u53e5\u5206\u522b\u7ecf\u8fc7LSTM\u795e \u7ecf\u7f51\u7edc\uff0c\u5f97\u5230\u6bcf\u4e2a\u5b50\u53e5\u7684\u7279\u5f81\u5411\u91cf\uff0c\u7136\u540e\u53d6\u8fd9\u4e9b\u7279\u5f81\u5411\u91cf\u7684\u5e73\u5747\u503c\uff0c\u4f5c\u4e3a\u4e2d\u95f4\u8bed\u4e49\u5411\u91cfC\u3002\u6211 \u4eec\u53ef\u4ee5\u5728\u4e24\u4e2a\u5c42\u6b21\u4e0a\u8ba8\u8bba\u8bdd\u8bed\u5e8f\u5217\uff1a\u6bcf\u7ec4\u5bf9\u8bdd\u662f\u7531\u5b50\u53e5\u5e8f\u5217\u7ec4\u6210\uff0c\u800c\u6bcf\u4e2a\u5b50\u53e5\u7531\u5355\u8bcd\u5e8f\u5217\u7ec4 \u6210\u3002\u57fa\u4e8e\u6b64\uff0c\u5bf9\u4e8e\u7ed9\u5b9a\u53e5\u5b50U m = {w m,1 , w m,2 , ..., w m,Nm }\uff0c\u7ecf\u8fc7LSTM\u7f51\u7edc\uff0c\u5c06\u6700\u540e\u4e00\u5c42\u7684\u9690 \u85cf\u5c42\u7684\u8f93\u51fah Nm \u4f5c\u4e3a\u8be5\u5b50\u53e5\u7684\u8bed\u4e49\u5411\u91cf\u3002\u672c\u6587\u4e3a\u5386\u53f2\u4fe1\u606f\u4e2d\u7684\u6bcf\u4e2a\u5b50\u53e5\u5206\u522b\u6784\u5efa\u4e00\u4e2aLSTM\u6a21 \u578b\uff0c\u53d6\u6240\u6709LSTM\u6a21\u578b\u8f93\u51fa\u7684\u5e73\u5747\u503c\u4f5c\u4e3a\u6574\u7ec4\u5bf9\u8bdd\u7684\u4e2d\u95f4\u8bed\u4e49\u5411\u91cfC\u3002 3.1.2 \u89e3 \u89e3 \u89e3\u7801 \u7801 \u7801\u5668 \u5668 \u5668 \u89e3\u7801\u5668\u662f\u5c06\u7f16\u7801\u5668\u4e2d\u751f\u6210\u7684\u4e2d\u95f4\u8bed\u4e49\u5411\u91cfC\u518d\u8f6c\u5316\u6210\u8f93\u51fa\u5e8f\u5217\u3002\u5728\u89e3\u7801\u9636\u6bb5\uff0c\u6211\u4eec\u4f1a\u7528\u8f93 \u51fa\u5e8f\u5217{y 1 , y 2 , ..., y t\u22121 }\u4ee5\u53ca\u4e2d\u95f4\u8bed\u4e49\u5411\u91cfC\u6765\u9884\u6d4b\u4e0b\u4e00\u4e2a\u8f93\u51fa\u7684\u5355\u8bcdy t ,\u5373 y t = argmax P (y t ) = T t=1 P (y t |y 1 , y 2 , ..., y t\u22121 , C) (2) \u901a\u5e38\u60c5\u51b5\u4e0b\uff0c\u4f1a\u4e3a\u89e3\u7801\u5668\u6307\u5b9a\u751f\u6210\u6587\u672c\u6700\u5927\u957f\u5ea6\u548c\u7ed3\u675f\u5b57\u7b26\u3002\u5728\u89e3\u7801\u8fc7\u7a0b\u4e2d\uff0c\u53ea\u8981\u7b26\u5408\u4e0a \u8ff0\u4e24\u4e2a\u6761\u4ef6\u4e4b\u4e00\uff0c\u89e3\u7801\u8fc7\u7a0b\u5c31\u4f1a\u7ed3\u675f\u3002\u89e3\u7801\u5668\u7684\u8f93\u51fa\u5e76\u4e0d\u662f\u6587\u672c\uff0c\u800c\u662f\u4e00\u4e2a\u5411\u91cf\uff0c\u8fd9\u4e2a\u5411\u91cf\u4ee3 \u8868\u7740\u5f53\u524d\u8fd9\u4e2a\u795e\u7ecf\u5355\u5143\u8f93\u51fa\u5bf9\u5e94\u8bcd\u8868\u7684\u6982\u7387\u5206\u5e03\uff0c\u901a\u5e38\u9009\u62e9\u6982\u7387\u6700\u9ad8\u7684\u4f5c\u4e3a\u8f93\u51fa\uff0c\u4f46\u6982\u7387\u6700\u9ad8 \u7684\u5f80\u5f80\u662f\"\u597d\u7684\"\uff0c\"\u662f\u7684\"\u7b49\u5b89\u5168\u56de\u590d\u3002\u56e0\u6b64\uff0c\u6211\u4eec\u5f15\u5165\u4e86\u4e3b\u9898\u3001\u60c5\u611f\u3001\u610f\u56fe\u8bc6\u522b\u5e2e\u52a9\u6a21\u578b\u66f4\u597d \u7684\u7406\u89e3\u5bf9\u8bdd\u5185\u5bb9\uff0c\u751f\u6210\u4e0e\u5bf9\u8bdd\u4e3b\u9898\u76f8\u5173\u4e14\u5177\u6709\u5408\u7406\u60c5\u611f\u548c\u610f\u56fe\u7684\u56de\u590d\uff0c\u589e\u52a0\u751f\u6210\u56de\u590d\u7684\u591a\u6837 \u6027\u3002 3.1.3 \u4e3b \u4e3b \u4e3b\u9898 \u9898 \u9898\u9884 \u9884 \u9884\u6d4b \u6d4b \u6d4b\u6a21 \u6a21 \u6a21\u578b \u578b \u578b \u4e3b\u9898\u9884\u6d4b\u4efb\u52a1\u65e8\u5728\u9884\u6d4b\u6574\u6bb5\u5bf9\u8bdd\u7684\u4e3b\u9898\u6807\u7b7e\uff0c\u7528\u4e8e\u7ea6\u675f\u751f\u6210\u56de\u590d\u7684\u4e3b\u9898\u3002\u672c\u5b9e\u9a8c\u4e2d\u53d6 \u5404\u4e2a\u5b50\u53e5\u5728\u7f16\u7801\u5668\u548c\u89e3\u7801\u5668\u4e2d\u7ecf\u8fc7LSTM\u7f51\u7edc\u5f97\u5230\u7684\u7279\u5f81\u5411\u91cf\u7684\u5e73\u5747\u503ch\u4f5c\u4e3a\u4e3b\u9898\u9884\u6d4b\u6a21\u578b \u7684softmax\u5c42\u7684\u8f93\u5165\u3002\u7ed9\u5b9a\u7279\u5f81\u5411\u91cfH t \u4f5c\u4e3asoftmax\u5c42\u7684\u8f93\u5165\uff1a P t s = softmax(W t s H t + B t s ) (3) \u8fd9\u91ccW t s \uff0cB t s \u4e3a\u6a21\u578b\u53c2\u6570\uff0cP t s \u662f\u610f\u56fe\u9884\u6d4b\u6a21\u578b\u7684\u8f93\u51fa\uff0c\u7528\u4e8e\u4e3b\u9898\u5206\u7c7b\u3002 3.1.4 \u60c5 \u60c5 \u60c5\u611f \u611f \u611f\u9884 \u9884 \u9884\u6d4b \u6d4b \u6d4b\u6a21 \u6a21 \u6a21\u578b \u578b \u578b \u60c5\u611f\u5206\u7c7b\u4efb\u52a1\u65e8\u5728\u9884\u6d4b\u5b50\u53e5\u7684\u610f\u56fe\u6807\u7b7e\uff0c\u7528\u4e8e\u7ea6\u675f\u751f\u6210\u56de\u590d\u7684\u60c5\u611f\u503e\u5411\u3002\u672c\u5b9e\u9a8c\u4e2d\u5c06\u6bcf\u4e2a \u5b50\u53e5\u5728\u7f16\u7801\u5668\u548c\u89e3\u7801\u5668\u4e2d\u7ecf\u8fc7LSTM\u7f51\u7edc\u5f97\u5230\u7684\u7279\u5f81\u5411\u91cf\u4f5c\u4e3a\u60c5\u611f\u9884\u6d4b\u6a21\u578b\u7684softmax\u5c42\u7684\u8f93 \u5165\u3002\u7ed9\u5b9a\u7279\u5f81\u5411\u91cfH s \u4f5c\u4e3asoftmax\u5c42\u7684\u8f93\u5165\uff1a P s s = softmax(W s s H s + B s s ) (4) \u8fd9\u91ccW s s \uff0cB s s \u4e3a\u6a21\u578b\u53c2\u6570\uff0cP s s \u662f\u60c5\u611f\u9884\u6d4b\u6a21\u578b\u7684\u8f93\u51fa\uff0c\u7528\u4e8e\u60c5\u611f\u5206\u7c7b\u3002 3.1.5 \u610f \u610f \u610f\u56fe \u56fe \u56fe\u9884 \u9884 \u9884\u6d4b \u6d4b \u6d4b\u6a21 \u6a21 \u6a21\u578b \u578b \u578b \u610f\u56fe\u5206\u7c7b\u4efb\u52a1\u65e8\u5728\u9884\u6d4b\u5b50\u53e5\u7684\u610f\u56fe\u6807\u7b7e\uff0c\u7528\u4e8e\u7ea6\u675f\u751f\u6210\u56de\u590d\u7684\u610f\u56fe\u503e\u5411\u3002\u672c\u5b9e\u9a8c\u4e2d\u5c06\u6bcf\u4e2a \u5b50\u53e5\u5728\u7f16\u7801\u5668\u548c\u89e3\u7801\u5668\u4e2d\u7ecf\u8fc7LSTM\u7f51\u7edc\u5f97\u5230\u7684\u7279\u5f81\u5411\u91cf\u4f5c\u4e3a\u610f\u56fe\u9884\u6d4b\u6a21\u578b\u7684softmax\u5c42\u7684\u8f93 \u5165\u3002\u7ed9\u5b9a\u7279\u5f81\u5411\u91cfH a \u4f5c\u4e3asoftmax\u5c42\u7684\u8f93\u5165\uff1a P a s = softmax(W a s H a + B a s )",
                        "eq_num": "(5)"
                    }
                ],
                "section": "",
                "sec_num": null
            },
            {
                "text": "\u3002Average\u8ba1 \u7b97 \u53e5 \u5b50 \u7ea7 \u522b \u7684 \u5d4c \u5165 \uff0c \u800cGreedy\u548cExtrema\u5219 \u8ba1 \u7b97 \u5355 \u8bcd \u7ea7 \u522b \u7684 \u4f59 \u5f26 \u76f8 \u4f3c \u5ea6 \u3002 \u5b83 \u4eec \u7684 \u533a \u522b \u5728 \u4e8e\uff0cGreedy\u5c06\u53e5\u5b50\u4e2d\u7684\u5e73\u5747\u5355\u8bcd\u5411\u91cf\u4f5c\u4e3a\u53e5\u5b50\u7684\u5d4c\u5165\uff0c\u800cExtrema\u5219\u91c7\u7528\u8fd9\u4e9b\u5355\u8bcd\u5411\u91cf\u7684\u6781 \u503c\u3002 4.2 \u5b9e \u5b9e \u5b9e\u9a8c \u9a8c \u9a8c\u53c2 \u53c2 \u53c2\u6570 \u6570 \u6570\u8bbe \u8bbe \u8bbe\u7f6e \u7f6e \u7f6e \u4e3a\u4e86\u83b7\u53d6\u6700\u4f18\u6a21\u578b\uff0c\u7ecf\u8c03\u6574\u672c\u5b9e\u9a8c\u8bbe\u7f6e\u7684\u76f8\u5173\u53c2\u6570\u5982\u8868 3\u6240\u793a\u3002 \u53c2 \u53c2 \u53c2\u6570 \u6570 \u6570 \u53d6 \u53d6 \u53d6\u503c \u503c \u503c Embedding\u5c42\u7684\u8f93\u51fa\u7ef4\u5ea6 128 LSTM\u5c42\u7684\u8f93\u51fa\u7ef4\u5ea6 128 \u8bdd\u8bed\u6700\u5927\u957f\u5ea6 20 \u6279\u6b21\u5927\u5c0f 64 \u8fed\u4ee3\u6b21\u6570 20 dropout 0.2 \u8868 3. DailyDialog\u57fa\u672c\u4fe1\u606f\u7edf\u8ba1 4.3 \u4e0e \u4e0e \u4e0e\u57fa \u57fa \u57fa\u51c6 \u51c6 \u51c6\u6a21 \u6a21 \u6a21\u578b \u578b \u578b\u6bd4 \u6bd4 \u6bd4\u8f83 \u8f83 \u8f83 \u5728 \u524d \u4eba \u5173 \u4e8e \u5bf9 \u8bdd \u56de \u590d \u751f \u6210 \u7684 \u7814 \u7a76 \u7684 \u57fa \u7840 \u4e0a \uff0c \u6211 \u4eec \u9009 \u53d6 \u4e86Seq2Seq\uff0cMulti\uff0cDir- VHRED\uff0cReCoSa\uff0cHRG\u4e94\u4e2a\u6a21\u578b\u8fdb\u884c\u5bf9\u6bd4\u5b9e\u9a8c\u3002\u63a5\u4e0b\u6765\uff0c\u6211\u4eec\u5c06\u5206\u522b\u4ecb\u7ecd\u8fd9\u4e94\u4e2a\u6a21\u578b\u3002 Seq2Seq\u6a21 \u6a21 \u6a21\u578b \u578b \u578b\uff1a \uff1a \uff1a (Bahdanau et al., 2015)\u5c06\u524d\u4e03\u53e5\u5386\u53f2\u5bf9\u8bdd\u4fe1\u606f\u62fc\u63a5\u6210\u4e00\u4e2a\u5411\u91cf\u4f5c\u4e3a\u8f93 \u5165\uff0c\u4e3a\u5176\u6784\u5efa\u4e00\u4e2aLSTM\u6a21\u578b\uff0c\u6b64LSTM\u6a21\u578b\u7684\u8f93\u51fa\u4f5c\u4e3a\u4e2d\u95f4\u8bed\u4e49\u5411\u91cf\uff0c\u7136\u540e\u89e3\u7801\u751f\u6210\u5bf9 \u8bdd\u56de\u590d\u3002\u6b64\u6a21\u578b\u4e3a\u4e00\u4e2a\u5355\u8f93\u5165\u6a21\u578b\u3002 Multi\u6a21 \u6a21 \u6a21\u578b \u578b \u578b\uff1a \uff1a \uff1a\u5c06\u524d\u4e03\u53e5\u8bdd\u5206\u522b\u4f5c\u4e3a\u8f93\u5165\uff0c\u4e3a\u6bcf\u4e2a\u5b50\u53e5\u6784\u5efaLSTM\u6a21\u578b\uff0c\u5c06\u6240\u6709LSTM\u7684\u8f93\u51fa \u7684\u5e73\u5747\u503c\u4f5c\u4e3a\u4e2d\u95f4\u8bed\u4e49\u5411\u91cf\uff0c\u7136\u540e\u89e3\u7801\u751f\u6210\u5bf9\u8bdd\u56de\u590d\u3002\u6b64\u6a21\u578b\u4e3a\u4e00\u4e2a\u591a\u8f93\u5165\u6a21\u578b\u3002 Dir-VHRED\u6a21 \u6a21 \u6a21 \u578b \u578b \u578b \uff1a \uff1a \uff1a (Zeng et al., 2019)\u4f7f \u7528Dirichlet\u5206 \u5e03 \u6765 \u63cf \u8ff0VHRED\u4e2d \u7684 \u6f5c \u5728 \u53d8 \u91cf\u3002Dirichlet\u5206\u5e03\u662f\u8d1d\u53f6\u65af\u7edf\u8ba1\u4e2d\u591a\u9879\u5f0f\u5206\u5e03\u7684\u6d41\u884c\u5171\u8f6d\u5148\u9a8c\uff0c\u5b83\u53ef\u4ee5\u662f\u51f9\u9762\u6216\u51f8\u9762\uff0c \u5355\u8c03\u4e0a\u5347\u3002",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "",
                "sec_num": null
            },
            {
                "text": "ReCoSa\u6a21 \u6a21 \u6a21\u578b \u578b \u578b\uff1a \uff1a \uff1a \u4ece\u7b2c\u4e00\u7ec4\u793a\u4f8b\u4e2d\u6211\u4eec\u53ef\u4ee5\u770b\u51fa\uff0c\u6b64\u8f6e\u5bf9\u8bdd\u7684\u4e3b\u9898\u662f\u6821\u56ed\u751f\u6d3b\u3002\u5bf9\u4e8e\u6700\u540e\u8bf4\u8bdd\u8005A\u7684\u63d0 \u95ee\uff0cMulti\u6a21\u578b\u6ca1\u80fd\u5145\u5206\u7684\u7406\u89e3\u8be5\u7ec4\u5bf9\u8bdd\u7684\u8bed\u4e49\u4fe1\u606f\uff0c\u6240\u4ee5\u751f\u6210\u4e86\"\u4f60\u53ef\u4ee5\u786e\u5b9a\"\u8fd9\u6837\u548c\u5bf9\u8bdd\u5185 \u5bb9\u6beb\u65e0\u8054\u7cfb\u7684\u56de\u590d\u3002\u800c\u6211\u4eec\u7684\u6a21\u578b\u56e0\u4e3a\u52a0\u5165\u4e86\u5bf9\u4e3b\u9898\u7684\u8bc6\u522b\uff0c\u6240\u4ee5\u751f\u6210\u4e86\u4e0e\u5bf9\u8bdd\u4e3b\u9898\u9ad8\u5ea6\u76f8\u5173 \u4e14\u81ea\u7136\u5408\u7406\u7684\u56de\u590d\u3002\u5728\u7b2c\u4e8c\u7ec4\u5bf9\u8bdd\u4e2d\uff0c\u6211\u4eec\u53d1\u73b0\u6700\u540e\u8bf4\u8bdd\u8005A\u7684\u610f\u56fe\u662f\"\u6307\u793a\"\u3002\u5728\u540c\u4e00\u8f6e\u5bf9\u8bdd \u4e2d\uff0c\u5f53\u524d\u4e00\u53e5\u7684\u610f\u56fe\u4e3a\"\u6307\u793a\"\u65f6\uff0c\u540e\u4e00\u53e5\u7684\u610f\u56fe\u6700\u5927\u6982\u7387\u4e3a\"\u627f\u8bfa\"\u3002\u6211\u4eec\u7684\u6a21\u578b\u56e0\u4e3a\u52a0\u5165\u4e86\u5bf9 \u610f\u56fe\u7684\u8bc6\u522b\uff0c\u751f\u6210\u4e86\u610f\u56fe\u4e3a\"\u627f\u8bfa\"\u7684\u56de\u7b54\uff0c\u800cmulti\u6a21\u578b\u5219\u751f\u6210\u4e86\u610f\u56fe\u4e3a\"\u9648\u8ff0\"\u7684\u56de\u7b54\uff0c\u663e\u7136\u6211 \u4eec\u751f\u6210\u7684\u56de\u590d\u66f4\u52a0\u7684\u5408\u7406\u3002\u5728\u7b2c\u4e09\u7ec4\u5b9e\u4f8b\u4e2d\uff0c\u6211\u4eec\u7684\u6a21\u578b\u751f\u6210\u7684\u56de\u590d\u5728\u610f\u56fe\u4e0a\u663e\u7136\u66f4\u52a0\u5408\u7406\u3002 \u53e6\u5916\uff0c\u6839\u636e4.1\u5c0f\u8282\u4e2d\u7684\u5206\u6790\uff0c\u540c\u4e00\u8bf4\u8bdd\u8005\u7684\u60c5\u611f\u57fa\u8c03\u5f80\u5f80\u662f\u4e0d\u53d8\u7684\u3002\u6211\u4eec\u7684\u6a21\u578b\u56e0\u4e3a\u52a0\u5165\u4e86\u5bf9 \u60c5\u611f\u7684\u8bc6\u522b\uff0c\u751f\u6210\u4e86\u5e26\u6709\u79ef\u6781\u60c5\u611f\u503e\u5411\u7684\u56de\u590d\u3002 5 \u603b \u603b \u603b\u7ed3 \u7ed3 \u7ed3",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "",
                "sec_num": null
            },
            {
                "text": "\u5728\u5927\u6570\u636e\u4e0d\u65ad\u53d1\u5c55\u7684\u4eca\u5929\uff0c\u4eba\u673a\u5bf9\u8bdd\u7cfb\u7edf\u662f\u4eba\u673a\u4ea4\u4e92\u9886\u57df\u4e00\u4e2a\u975e\u5e38\u91cd\u8981\u7684\u7814\u7a76\u65b9\u5411\uff0c\u5f00\u653e \u57df\u804a\u5929\u673a\u5668\u4eba\u7684\u7814\u7a76\u53d7\u5230\u4e86\u5e7f\u6cdb\u5173\u6ce8\u3002\u8fd9\u4f7f\u5f97\u63d0\u9ad8\u5bf9\u8bdd\u56de\u590d\u7684\u8d28\u91cf\u5c24\u4e3a\u91cd\u8981\u3002 \u8003 \u8651 \u5230 \u5728 \u5bf9 \u8bdd \u5f80 \u5f80 \u662f \u56f4 \u7ed5 \u7279 \u5b9a \u7684 \u4e3b \u9898 \u4e14 \u5bf9 \u8bdd \u4e2d \u7684 \u6bcf \u4e2a \u5b50 \u53e5 \u5177 \u6709 \u660e \u663e \u7684 \u60c5 \u611f \u548c \u610f \u56fe \u503e \u5411\uff0c\u672c\u6587\u63d0\u51fa\u4e86\u4e00\u79cd\u57fa\u4e8e\u5bf9\u8bdd\u7ea6\u675f\u7684\u56de\u590d\u751f\u6210\u6a21\u578b\u3002\u4ee5\u5f80\u7684\u5bf9\u8bdd\u751f\u6210\u6a21\u578b\u603b\u662f\u503e\u5411\u4e8e\u751f \u6210\u7c7b\u4f3c\"\u597d\u7684\"\u3001\"\u6211\u4e0d\u77e5\u9053\"\u7684\u5b89\u5168\u56de\u590d\uff0c\u867d\u7136\u8fd9\u6837\u7684\u56de\u590d\u770b\u8d77\u6765\u5408\u7406\uff0c\u4f46\u5176\u5b9e\u8fd9\u6837\u7684\u6a21\u7cca \u56de \u590d \u610f \u4e49 \u5e76 \u4e0d \u5927 \uff0c \u4fe1 \u606f \u91cf \u5f88 \u4f4e \u3002 \u4e3a \u4e86 \u89e3 \u51b3 \u8fd9 \u4e2a \u95ee \u9898 \uff0c \u6211 \u4eec \u5728 \u57fa \u4e8eLSTM\u7f51 \u7edc \u7684Seq2Seq\u6a21 \u578b \u57fa \u7840 \u4e0a \u5f15 \u5165 \u4e86 \u5bf9 \u5bf9 \u8bdd \u7684 \u8bdd \u9898 \u3001 \u60c5 \u611f \u548c \u610f \u56fe \u7684 \u8bc6 \u522b \uff0c \u4f7f \u5f97 \u751f \u6210 \u5177 \u6709 \u5408 \u7406 \u7684 \u60c5 \u611f \u548c \u610f",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "",
                "sec_num": null
            }
        ],
        "back_matter": [
            {
                "text": "A \u6211\u5982\u4f55\u7533\u8bf7\u7f8e\u56fd\u7684\u5927\u5b66\uff1f Kyunghyun Cho, and Yoshua Bengio. 2015 . Neural machine translation by jointly learning to align and translate. In Yoshua Bengio and Yann LeCun, editors, 3rd International Conference on Learning Representations, ICLR 2015, San Diego, CA, USA, May 7-9, 2015, Conference Track Proceedings.",
                "cite_spans": [
                    {
                        "start": 14,
                        "end": 52,
                        "text": "Kyunghyun Cho, and Yoshua Bengio. 2015",
                        "ref_id": null
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "\u793a \u793a \u793a\u4f8b \u4f8b \u4f8b1",
                "sec_num": null
            }
        ],
        "bib_entries": {
            "BIBREF0": {
                "ref_id": "b0",
                "title": "The mathematics of statistical machine translation: Parameter estimation",
                "authors": [
                    {
                        "first": "F",
                        "middle": [],
                        "last": "Peter",
                        "suffix": ""
                    },
                    {
                        "first": "Stephen",
                        "middle": [
                            "Della"
                        ],
                        "last": "Brown",
                        "suffix": ""
                    },
                    {
                        "first": "Vincent",
                        "middle": [
                            "J"
                        ],
                        "last": "Pietra",
                        "suffix": ""
                    },
                    {
                        "first": "Robert",
                        "middle": [
                            "L"
                        ],
                        "last": "Della Pietra",
                        "suffix": ""
                    },
                    {
                        "first": "",
                        "middle": [],
                        "last": "Mercer",
                        "suffix": ""
                    }
                ],
                "year": 1993,
                "venue": "Comput. Linguistics",
                "volume": "19",
                "issue": "2",
                "pages": "263--311",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Peter F. Brown, Stephen Della Pietra, Vincent J. Della Pietra, and Robert L. Mercer. 1993. The mathe- matics of statistical machine translation: Parameter estimation. Comput. Linguistics, 19(2):263-311.",
                "links": null
            },
            "BIBREF1": {
                "ref_id": "b1",
                "title": "Learning phrase representations using RNN encoder-decoder for statistical machine translation",
                "authors": [
                    {
                        "first": "Kyunghyun",
                        "middle": [],
                        "last": "Cho",
                        "suffix": ""
                    },
                    {
                        "first": "Bart",
                        "middle": [],
                        "last": "Van Merrienboer",
                        "suffix": ""
                    },
                    {
                        "first": "Dzmitry",
                        "middle": [],
                        "last": "Aglar G\u00fcl\u00e7ehre",
                        "suffix": ""
                    },
                    {
                        "first": "Fethi",
                        "middle": [],
                        "last": "Bahdanau",
                        "suffix": ""
                    },
                    {
                        "first": "Holger",
                        "middle": [],
                        "last": "Bougares",
                        "suffix": ""
                    },
                    {
                        "first": "Yoshua",
                        "middle": [],
                        "last": "Schwenk",
                        "suffix": ""
                    },
                    {
                        "first": "",
                        "middle": [],
                        "last": "Bengio",
                        "suffix": ""
                    }
                ],
                "year": 2014,
                "venue": "Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing",
                "volume": "",
                "issue": "",
                "pages": "1724--1734",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Kyunghyun Cho, Bart van Merrienboer, \u00c7 aglar G\u00fcl\u00e7ehre, Dzmitry Bahdanau, Fethi Bougares, Holger Schwenk, and Yoshua Bengio. 2014. Learning phrase representations using RNN encoder-decoder for statistical machine translation. In Alessandro Moschitti, Bo Pang, and Walter Daelemans, editors, Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing, EMNLP 2014, October 25-29, 2014, Doha, Qatar, A meeting of SIGDAT, a Special Interest Group of the ACL, pages 1724-1734. ACL.",
                "links": null
            },
            "BIBREF2": {
                "ref_id": "b2",
                "title": "The measurement of textual coherence with latent semantic analysis",
                "authors": [
                    {
                        "first": "W",
                        "middle": [],
                        "last": "Peter",
                        "suffix": ""
                    },
                    {
                        "first": "Walter",
                        "middle": [],
                        "last": "Foltz",
                        "suffix": ""
                    },
                    {
                        "first": "Thomas",
                        "middle": [
                            "K"
                        ],
                        "last": "Kintsch",
                        "suffix": ""
                    },
                    {
                        "first": "",
                        "middle": [],
                        "last": "Landauer",
                        "suffix": ""
                    }
                ],
                "year": 1998,
                "venue": "Discourse processes",
                "volume": "25",
                "issue": "",
                "pages": "285--307",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Peter W Foltz, Walter Kintsch, and Thomas K Landauer. 1998. The measurement of textual coherence with latent semantic analysis. Discourse processes, 25(2-3):285-307.",
                "links": null
            },
            "BIBREF3": {
                "ref_id": "b3",
                "title": "Bootstrapping dialog systems with word embeddings",
                "authors": [
                    {
                        "first": "Gabriel",
                        "middle": [],
                        "last": "Forgues",
                        "suffix": ""
                    },
                    {
                        "first": "Joelle",
                        "middle": [],
                        "last": "Pineau",
                        "suffix": ""
                    },
                    {
                        "first": "Jean-Marie",
                        "middle": [],
                        "last": "Larchev\u00eaque",
                        "suffix": ""
                    },
                    {
                        "first": "R\u00e9al",
                        "middle": [],
                        "last": "Tremblay",
                        "suffix": ""
                    }
                ],
                "year": 2014,
                "venue": "Nips, modern machine learning and natural language processing workshop",
                "volume": "2",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Gabriel Forgues, Joelle Pineau, Jean-Marie Larchev\u00eaque, and R\u00e9al Tremblay. 2014. Bootstrapping dialog systems with word embeddings. In Nips, modern machine learning and natural language processing workshop, volume 2.",
                "links": null
            },
            "BIBREF4": {
                "ref_id": "b4",
                "title": "Improving neural parsing by disentangling model combination and reranking effects",
                "authors": [
                    {
                        "first": "Daniel",
                        "middle": [],
                        "last": "Fried",
                        "suffix": ""
                    },
                    {
                        "first": "Mitchell",
                        "middle": [],
                        "last": "Stern",
                        "suffix": ""
                    },
                    {
                        "first": "Dan",
                        "middle": [],
                        "last": "Klein",
                        "suffix": ""
                    }
                ],
                "year": 2017,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {
                    "arXiv": [
                        "arXiv:1707.03058"
                    ]
                },
                "num": null,
                "urls": [],
                "raw_text": "Daniel Fried, Mitchell Stern, and Dan Klein. 2017. Improving neural parsing by disentangling model combination and reranking effects. arXiv preprint arXiv:1707.03058.",
                "links": null
            },
            "BIBREF5": {
                "ref_id": "b5",
                "title": "Improved training of wasserstein gans",
                "authors": [
                    {
                        "first": "Ishaan",
                        "middle": [],
                        "last": "Gulrajani",
                        "suffix": ""
                    },
                    {
                        "first": "Faruk",
                        "middle": [],
                        "last": "Ahmed",
                        "suffix": ""
                    },
                    {
                        "first": "Martin",
                        "middle": [],
                        "last": "Arjovsky",
                        "suffix": ""
                    },
                    {
                        "first": "Vincent",
                        "middle": [],
                        "last": "Dumoulin",
                        "suffix": ""
                    },
                    {
                        "first": "Aaron C",
                        "middle": [],
                        "last": "Courville",
                        "suffix": ""
                    }
                ],
                "year": 2017,
                "venue": "Advances in neural information processing systems",
                "volume": "",
                "issue": "",
                "pages": "5767--5777",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Ishaan Gulrajani, Faruk Ahmed, Martin Arjovsky, Vincent Dumoulin, and Aaron C Courville. 2017. Improved training of wasserstein gans. In Advances in neural information processing systems, pages 5767-5777.",
                "links": null
            },
            "BIBREF6": {
                "ref_id": "b6",
                "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
            },
            "BIBREF7": {
                "ref_id": "b7",
                "title": "Bidirectional LSTM-CRF models for sequence tagging",
                "authors": [
                    {
                        "first": "Zhiheng",
                        "middle": [],
                        "last": "Huang",
                        "suffix": ""
                    },
                    {
                        "first": "Wei",
                        "middle": [],
                        "last": "Xu",
                        "suffix": ""
                    },
                    {
                        "first": "Kai",
                        "middle": [],
                        "last": "Yu",
                        "suffix": ""
                    }
                ],
                "year": 2015,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Zhiheng Huang, Wei Xu, and Kai Yu. 2015. Bidirectional LSTM-CRF models for sequence tagging. CoRR, abs/1508.01991.",
                "links": null
            },
            "BIBREF8": {
                "ref_id": "b8",
                "title": "Ask me anything: Dynamic memory networks for natural language processing",
                "authors": [
                    {
                        "first": "Ankit",
                        "middle": [],
                        "last": "Kumar",
                        "suffix": ""
                    },
                    {
                        "first": "Ozan",
                        "middle": [],
                        "last": "Irsoy",
                        "suffix": ""
                    },
                    {
                        "first": "Peter",
                        "middle": [],
                        "last": "Ondruska",
                        "suffix": ""
                    },
                    {
                        "first": "Mohit",
                        "middle": [],
                        "last": "Iyyer",
                        "suffix": ""
                    },
                    {
                        "first": "James",
                        "middle": [],
                        "last": "Bradbury",
                        "suffix": ""
                    },
                    {
                        "first": "Ishaan",
                        "middle": [],
                        "last": "Gulrajani",
                        "suffix": ""
                    },
                    {
                        "first": "Victor",
                        "middle": [],
                        "last": "Zhong",
                        "suffix": ""
                    },
                    {
                        "first": "Romain",
                        "middle": [],
                        "last": "Paulus",
                        "suffix": ""
                    },
                    {
                        "first": "Richard",
                        "middle": [],
                        "last": "Socher",
                        "suffix": ""
                    }
                ],
                "year": 2016,
                "venue": "International conference on machine learning",
                "volume": "",
                "issue": "",
                "pages": "1378--1387",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Ankit Kumar, Ozan Irsoy, Peter Ondruska, Mohit Iyyer, James Bradbury, Ishaan Gulrajani, Victor Zhong, Romain Paulus, and Richard Socher. 2016. Ask me anything: Dynamic memory networks for natural language processing. In International conference on machine learning, pages 1378-1387.",
                "links": null
            },
            "BIBREF9": {
                "ref_id": "b9",
                "title": "Deep learning. nature",
                "authors": [
                    {
                        "first": "Yann",
                        "middle": [],
                        "last": "Lecun",
                        "suffix": ""
                    },
                    {
                        "first": "Yoshua",
                        "middle": [],
                        "last": "Bengio",
                        "suffix": ""
                    },
                    {
                        "first": "Geoffrey",
                        "middle": [],
                        "last": "Hinton",
                        "suffix": ""
                    }
                ],
                "year": 2015,
                "venue": "",
                "volume": "521",
                "issue": "",
                "pages": "436--444",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Yann LeCun, Yoshua Bengio, and Geoffrey Hinton. 2015. Deep learning. nature, 521(7553):436-444.",
                "links": null
            },
            "BIBREF10": {
                "ref_id": "b10",
                "title": "A diversity-promoting objective function for neural conversation models",
                "authors": [
                    {
                        "first": "Jiwei",
                        "middle": [],
                        "last": "Li",
                        "suffix": ""
                    },
                    {
                        "first": "Michel",
                        "middle": [],
                        "last": "Galley",
                        "suffix": ""
                    },
                    {
                        "first": "Chris",
                        "middle": [],
                        "last": "Brockett",
                        "suffix": ""
                    },
                    {
                        "first": "Jianfeng",
                        "middle": [],
                        "last": "Gao",
                        "suffix": ""
                    },
                    {
                        "first": "Bill",
                        "middle": [],
                        "last": "Dolan",
                        "suffix": ""
                    }
                ],
                "year": 2016,
                "venue": "The 2016 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies",
                "volume": "",
                "issue": "",
                "pages": "110--119",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Jiwei Li, Michel Galley, Chris Brockett, Jianfeng Gao, and Bill Dolan. 2016. A diversity-promoting ob- jective function for neural conversation models. In Kevin Knight, Ani Nenkova, and Owen Rambow, editors, NAACL HLT 2016, The 2016 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, San Diego California, USA, June 12-17, 2016, pages 110-119. The Association for Computational Linguistics.",
                "links": null
            },
            "BIBREF11": {
                "ref_id": "b11",
                "title": "Adversarial learning for neural dialogue generation",
                "authors": [
                    {
                        "first": "Jiwei",
                        "middle": [],
                        "last": "Li",
                        "suffix": ""
                    },
                    {
                        "first": "Will",
                        "middle": [],
                        "last": "Monroe",
                        "suffix": ""
                    },
                    {
                        "first": "Tianlin",
                        "middle": [],
                        "last": "Shi",
                        "suffix": ""
                    },
                    {
                        "first": "S\u00e9bastien",
                        "middle": [],
                        "last": "Jean",
                        "suffix": ""
                    },
                    {
                        "first": "Alan",
                        "middle": [],
                        "last": "Ritter",
                        "suffix": ""
                    },
                    {
                        "first": "Dan",
                        "middle": [],
                        "last": "Jurafsky",
                        "suffix": ""
                    }
                ],
                "year": 2017,
                "venue": "Proceedings of the 2017 Conference on Empirical Methods in Natural Language Processing",
                "volume": "",
                "issue": "",
                "pages": "2157--2169",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Jiwei Li, Will Monroe, Tianlin Shi, S\u00e9bastien Jean, Alan Ritter, and Dan Jurafsky. 2017a. Adversarial learning for neural dialogue generation. In Martha Palmer, Rebecca Hwa, and Sebastian Riedel, editors, Proceedings of the 2017 Conference on Empirical Methods in Natural Language Processing, EMNLP 2017, Copenhagen, Denmark, September 9-11, 2017, pages 2157-2169. Association for Computational Linguistics.",
                "links": null
            },
            "BIBREF12": {
                "ref_id": "b12",
                "title": "Dailydialog: A manually labelled multi-turn dialogue dataset",
                "authors": [
                    {
                        "first": "Yanran",
                        "middle": [],
                        "last": "Li",
                        "suffix": ""
                    },
                    {
                        "first": "Hui",
                        "middle": [],
                        "last": "Su",
                        "suffix": ""
                    },
                    {
                        "first": "Xiaoyu",
                        "middle": [],
                        "last": "Shen",
                        "suffix": ""
                    },
                    {
                        "first": "Wenjie",
                        "middle": [],
                        "last": "Li",
                        "suffix": ""
                    },
                    {
                        "first": "Ziqiang",
                        "middle": [],
                        "last": "Cao",
                        "suffix": ""
                    },
                    {
                        "first": "Shuzi",
                        "middle": [],
                        "last": "Niu",
                        "suffix": ""
                    }
                ],
                "year": 2017,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {
                    "arXiv": [
                        "arXiv:1710.03957"
                    ]
                },
                "num": null,
                "urls": [],
                "raw_text": "Yanran Li, Hui Su, Xiaoyu Shen, Wenjie Li, Ziqiang Cao, and Shuzi Niu. 2017b. Dailydialog: A manually labelled multi-turn dialogue dataset. arXiv preprint arXiv:1710.03957.",
                "links": null
            },
            "BIBREF13": {
                "ref_id": "b13",
                "title": "Practical graph isomorphism",
                "authors": [
                    {
                        "first": "D",
                        "middle": [],
                        "last": "Brendan",
                        "suffix": ""
                    },
                    {
                        "first": "Adolfo",
                        "middle": [],
                        "last": "Mckay",
                        "suffix": ""
                    },
                    {
                        "first": "",
                        "middle": [],
                        "last": "Piperno",
                        "suffix": ""
                    }
                ],
                "year": 2014,
                "venue": "II. J. Symb. Comput",
                "volume": "60",
                "issue": "",
                "pages": "94--112",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Brendan D. McKay and Adolfo Piperno. 2014. Practical graph isomorphism, II. J. Symb. Comput., 60:94-112.",
                "links": null
            },
            "BIBREF14": {
                "ref_id": "b14",
                "title": "Sequence to backward and forward sequences: A content-introducing approach to generative short-text conversation",
                "authors": [
                    {
                        "first": "Lili",
                        "middle": [],
                        "last": "Mou",
                        "suffix": ""
                    },
                    {
                        "first": "Yiping",
                        "middle": [],
                        "last": "Song",
                        "suffix": ""
                    },
                    {
                        "first": "Rui",
                        "middle": [],
                        "last": "Yan",
                        "suffix": ""
                    },
                    {
                        "first": "Ge",
                        "middle": [],
                        "last": "Li",
                        "suffix": ""
                    },
                    {
                        "first": "Lu",
                        "middle": [],
                        "last": "Zhang",
                        "suffix": ""
                    },
                    {
                        "first": "Zhi",
                        "middle": [],
                        "last": "Jin",
                        "suffix": ""
                    }
                ],
                "year": 2016,
                "venue": "COLING 2016, 26th International Conference on Computational Linguistics, Proceedings of the Conference: Technical Papers",
                "volume": "",
                "issue": "",
                "pages": "3349--3358",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Lili Mou, Yiping Song, Rui Yan, Ge Li, Lu Zhang, and Zhi Jin. 2016. Sequence to backward and forward sequences: A content-introducing approach to generative short-text conversation. In Nicoletta Cal- zolari, Yuji Matsumoto, and Rashmi Prasad, editors, COLING 2016, 26th International Conference on Computational Linguistics, Proceedings of the Conference: Technical Papers, December 11-16, 2016, Osaka, Japan, pages 3349-3358. ACL.",
                "links": null
            },
            "BIBREF15": {
                "ref_id": "b15",
                "title": "Bleu: a method for automatic evaluation of machine translation",
                "authors": [
                    {
                        "first": "Kishore",
                        "middle": [],
                        "last": "Papineni",
                        "suffix": ""
                    },
                    {
                        "first": "Salim",
                        "middle": [],
                        "last": "Roukos",
                        "suffix": ""
                    },
                    {
                        "first": "Todd",
                        "middle": [],
                        "last": "Ward",
                        "suffix": ""
                    },
                    {
                        "first": "Wei-Jing",
                        "middle": [],
                        "last": "Zhu",
                        "suffix": ""
                    }
                ],
                "year": 2002,
                "venue": "Proceedings of the 40th annual meeting on association for computational linguistics",
                "volume": "",
                "issue": "",
                "pages": "311--318",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Kishore Papineni, Salim Roukos, Todd Ward, and Wei-Jing Zhu. 2002. Bleu: a method for automatic evaluation of machine translation. In Proceedings of the 40th annual meeting on association for computational linguistics, pages 311-318. Association for Computational Linguistics.",
                "links": null
            },
            "BIBREF16": {
                "ref_id": "b16",
                "title": "An optimal assessment of natural language student input using word-to-word similarity metrics",
                "authors": [
                    {
                        "first": "Vasile",
                        "middle": [],
                        "last": "Rus",
                        "suffix": ""
                    },
                    {
                        "first": "Mihai",
                        "middle": [],
                        "last": "Lintean",
                        "suffix": ""
                    }
                ],
                "year": 2012,
                "venue": "International Conference on Intelligent Tutoring Systems",
                "volume": "",
                "issue": "",
                "pages": "675--676",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Vasile Rus and Mihai Lintean. 2012. An optimal assessment of natural language student input using word-to-word similarity metrics. In International Conference on Intelligent Tutoring Systems, pages 675-676. Springer.",
                "links": null
            },
            "BIBREF17": {
                "ref_id": "b17",
                "title": "Building end-to-end dialogue systems using generative hierarchical neural network models",
                "authors": [
                    {
                        "first": "Alessandro",
                        "middle": [],
                        "last": "Iulian V Serban",
                        "suffix": ""
                    },
                    {
                        "first": "Yoshua",
                        "middle": [],
                        "last": "Sordoni",
                        "suffix": ""
                    },
                    {
                        "first": "Aaron",
                        "middle": [],
                        "last": "Bengio",
                        "suffix": ""
                    },
                    {
                        "first": "Joelle",
                        "middle": [],
                        "last": "Courville",
                        "suffix": ""
                    },
                    {
                        "first": "",
                        "middle": [],
                        "last": "Pineau",
                        "suffix": ""
                    }
                ],
                "year": 2016,
                "venue": "Thirtieth AAAI Conference on Artificial Intelligence",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Iulian V Serban, Alessandro Sordoni, Yoshua Bengio, Aaron Courville, and Joelle Pineau. 2016. Building end-to-end dialogue systems using generative hierarchical neural network models. In Thirtieth AAAI Conference on Artificial Intelligence.",
                "links": null
            },
            "BIBREF18": {
                "ref_id": "b18",
                "title": "Generating high-quality and informative conversation responses with sequence-to-sequence models",
                "authors": [
                    {
                        "first": "Louis",
                        "middle": [],
                        "last": "Shao",
                        "suffix": ""
                    },
                    {
                        "first": "Stephan",
                        "middle": [],
                        "last": "Gouws",
                        "suffix": ""
                    },
                    {
                        "first": "Denny",
                        "middle": [],
                        "last": "Britz",
                        "suffix": ""
                    },
                    {
                        "first": "Anna",
                        "middle": [],
                        "last": "Goldie",
                        "suffix": ""
                    },
                    {
                        "first": "Brian",
                        "middle": [],
                        "last": "Strope",
                        "suffix": ""
                    },
                    {
                        "first": "Ray",
                        "middle": [],
                        "last": "Kurzweil",
                        "suffix": ""
                    }
                ],
                "year": 2017,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {
                    "arXiv": [
                        "arXiv:1701.03185"
                    ]
                },
                "num": null,
                "urls": [],
                "raw_text": "Louis Shao, Stephan Gouws, Denny Britz, Anna Goldie, Brian Strope, and Ray Kurzweil. 2017. Gener- ating high-quality and informative conversation responses with sequence-to-sequence models. arXiv preprint arXiv:1701.03185.",
                "links": null
            },
            "BIBREF19": {
                "ref_id": "b19",
                "title": "A neural network approach to context-sensitive generation of conversational responses",
                "authors": [
                    {
                        "first": "Alessandro",
                        "middle": [],
                        "last": "Sordoni",
                        "suffix": ""
                    },
                    {
                        "first": "Michel",
                        "middle": [],
                        "last": "Galley",
                        "suffix": ""
                    },
                    {
                        "first": "Michael",
                        "middle": [],
                        "last": "Auli",
                        "suffix": ""
                    },
                    {
                        "first": "Chris",
                        "middle": [],
                        "last": "Brockett",
                        "suffix": ""
                    },
                    {
                        "first": "Yangfeng",
                        "middle": [],
                        "last": "Ji",
                        "suffix": ""
                    },
                    {
                        "first": "Margaret",
                        "middle": [],
                        "last": "Mitchell",
                        "suffix": ""
                    },
                    {
                        "first": "Jian-Yun",
                        "middle": [],
                        "last": "Nie",
                        "suffix": ""
                    },
                    {
                        "first": "Jianfeng",
                        "middle": [],
                        "last": "Gao",
                        "suffix": ""
                    },
                    {
                        "first": "Bill",
                        "middle": [],
                        "last": "Dolan",
                        "suffix": ""
                    }
                ],
                "year": 2015,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {
                    "arXiv": [
                        "arXiv:1506.06714"
                    ]
                },
                "num": null,
                "urls": [],
                "raw_text": "Alessandro Sordoni, Michel Galley, Michael Auli, Chris Brockett, Yangfeng Ji, Margaret Mitchell, Jian- Yun Nie, Jianfeng Gao, and Bill Dolan. 2015. A neural network approach to context-sensitive generation of conversational responses. arXiv preprint arXiv:1506.06714.",
                "links": null
            },
            "BIBREF20": {
                "ref_id": "b20",
                "title": "Sequence to sequence learning with neural networks",
                "authors": [
                    {
                        "first": "Ilya",
                        "middle": [],
                        "last": "Sutskever",
                        "suffix": ""
                    },
                    {
                        "first": "Oriol",
                        "middle": [],
                        "last": "Vinyals",
                        "suffix": ""
                    },
                    {
                        "first": "V",
                        "middle": [],
                        "last": "Quoc",
                        "suffix": ""
                    },
                    {
                        "first": "",
                        "middle": [],
                        "last": "Le",
                        "suffix": ""
                    }
                ],
                "year": 2014,
                "venue": "Advances in Neural Information Processing Systems 27: Annual Conference on Neural Information Processing Systems",
                "volume": "",
                "issue": "",
                "pages": "3104--3112",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Ilya Sutskever, Oriol Vinyals, and Quoc V. Le. 2014. Sequence to sequence learning with neural networks. In Zoubin Ghahramani, Max Welling, Corinna Cortes, Neil D. Lawrence, and Kilian Q. Weinberger, editors, Advances in Neural Information Processing Systems 27: Annual Conference on Neural Information Processing Systems 2014, December 8-13 2014, Montreal, Quebec, Canada, pages 3104- 3112.",
                "links": null
            },
            "BIBREF21": {
                "ref_id": "b21",
                "title": "Get the point of my utterance! learning towards effective responses with multi-head attention mechanism",
                "authors": [
                    {
                        "first": "Chongyang",
                        "middle": [],
                        "last": "Tao",
                        "suffix": ""
                    },
                    {
                        "first": "Shen",
                        "middle": [],
                        "last": "Gao",
                        "suffix": ""
                    },
                    {
                        "first": "Mingyue",
                        "middle": [],
                        "last": "Shang",
                        "suffix": ""
                    },
                    {
                        "first": "Wei",
                        "middle": [],
                        "last": "Wu",
                        "suffix": ""
                    },
                    {
                        "first": "Dongyan",
                        "middle": [],
                        "last": "Zhao",
                        "suffix": ""
                    },
                    {
                        "first": "Rui",
                        "middle": [],
                        "last": "Yan",
                        "suffix": ""
                    }
                ],
                "year": 2018,
                "venue": "IJCAI",
                "volume": "",
                "issue": "",
                "pages": "4418--4424",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Chongyang Tao, Shen Gao, Mingyue Shang, Wei Wu, Dongyan Zhao, and Rui Yan. 2018. Get the point of my utterance! learning towards effective responses with multi-head attention mechanism. In IJCAI, pages 4418-4424.",
                "links": null
            },
            "BIBREF22": {
                "ref_id": "b22",
                "title": "Diverse beam search: Decoding diverse solutions from neural sequence models",
                "authors": [
                    {
                        "first": "K",
                        "middle": [],
                        "last": "Ashwin",
                        "suffix": ""
                    },
                    {
                        "first": "Michael",
                        "middle": [],
                        "last": "Vijayakumar",
                        "suffix": ""
                    },
                    {
                        "first": "Ramprasaath",
                        "middle": [
                            "R"
                        ],
                        "last": "Cogswell",
                        "suffix": ""
                    },
                    {
                        "first": "Qing",
                        "middle": [],
                        "last": "Selvaraju",
                        "suffix": ""
                    },
                    {
                        "first": "Stefan",
                        "middle": [],
                        "last": "Sun",
                        "suffix": ""
                    },
                    {
                        "first": "David",
                        "middle": [
                            "J"
                        ],
                        "last": "Lee",
                        "suffix": ""
                    },
                    {
                        "first": "Dhruv",
                        "middle": [],
                        "last": "Crandall",
                        "suffix": ""
                    },
                    {
                        "first": "",
                        "middle": [],
                        "last": "Batra",
                        "suffix": ""
                    }
                ],
                "year": 2016,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Ashwin K. Vijayakumar, Michael Cogswell, Ramprasaath R. Selvaraju, Qing Sun, Stefan Lee, David J. Crandall, and Dhruv Batra. 2016. Diverse beam search: Decoding diverse solutions from neural sequence models. CoRR, abs/1610.02424.",
                "links": null
            },
            "BIBREF23": {
                "ref_id": "b23",
                "title": "A neural conversational model",
                "authors": [
                    {
                        "first": "Oriol",
                        "middle": [],
                        "last": "Vinyals",
                        "suffix": ""
                    },
                    {
                        "first": "V",
                        "middle": [],
                        "last": "Quoc",
                        "suffix": ""
                    },
                    {
                        "first": "",
                        "middle": [],
                        "last": "Le",
                        "suffix": ""
                    }
                ],
                "year": 2015,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Oriol Vinyals and Quoc V. Le. 2015. A neural conversational model. CoRR, abs/1506.05869.",
                "links": null
            },
            "BIBREF24": {
                "ref_id": "b24",
                "title": "Semantically conditioned lstm-based natural language generation for spoken dialogue systems",
                "authors": [
                    {
                        "first": "Milica",
                        "middle": [],
                        "last": "Tsung-Hsien Wen",
                        "suffix": ""
                    },
                    {
                        "first": "Nikola",
                        "middle": [],
                        "last": "Gasic",
                        "suffix": ""
                    },
                    {
                        "first": "Pei-Hao",
                        "middle": [],
                        "last": "Mrksic",
                        "suffix": ""
                    },
                    {
                        "first": "David",
                        "middle": [],
                        "last": "Su",
                        "suffix": ""
                    },
                    {
                        "first": "Steve",
                        "middle": [],
                        "last": "Vandyke",
                        "suffix": ""
                    },
                    {
                        "first": "",
                        "middle": [],
                        "last": "Young",
                        "suffix": ""
                    }
                ],
                "year": 2015,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {
                    "arXiv": [
                        "arXiv:1508.01745"
                    ]
                },
                "num": null,
                "urls": [],
                "raw_text": "Tsung-Hsien Wen, Milica Gasic, Nikola Mrksic, Pei-Hao Su, David Vandyke, and Steve Young. 2015. Semantically conditioned lstm-based natural language generation for spoken dialogue systems. arXiv preprint arXiv:1508.01745.",
                "links": null
            },
            "BIBREF25": {
                "ref_id": "b25",
                "title": "Topic aware neural response generation",
                "authors": [
                    {
                        "first": "Chen",
                        "middle": [],
                        "last": "Xing",
                        "suffix": ""
                    },
                    {
                        "first": "Wei",
                        "middle": [],
                        "last": "Wu",
                        "suffix": ""
                    },
                    {
                        "first": "Yu",
                        "middle": [],
                        "last": "Wu",
                        "suffix": ""
                    },
                    {
                        "first": "Jie",
                        "middle": [],
                        "last": "Liu",
                        "suffix": ""
                    },
                    {
                        "first": "Yalou",
                        "middle": [],
                        "last": "Huang",
                        "suffix": ""
                    },
                    {
                        "first": "Ming",
                        "middle": [],
                        "last": "Zhou",
                        "suffix": ""
                    },
                    {
                        "first": "Wei-Ying",
                        "middle": [],
                        "last": "Ma",
                        "suffix": ""
                    }
                ],
                "year": 2017,
                "venue": "Thirty-First AAAI Conference on Artificial Intelligence",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Chen Xing, Wei Wu, Yu Wu, Jie Liu, Yalou Huang, Ming Zhou, and Wei-Ying Ma. 2017. Topic aware neural response generation. In Thirty-First AAAI Conference on Artificial Intelligence.",
                "links": null
            },
            "BIBREF26": {
                "ref_id": "b26",
                "title": "Neural contextual conversation learning with labeled question-answering pairs",
                "authors": [
                    {
                        "first": "Kun",
                        "middle": [],
                        "last": "Xiong",
                        "suffix": ""
                    },
                    {
                        "first": "Anqi",
                        "middle": [],
                        "last": "Cui",
                        "suffix": ""
                    },
                    {
                        "first": "Zefeng",
                        "middle": [],
                        "last": "Zhang",
                        "suffix": ""
                    },
                    {
                        "first": "Ming",
                        "middle": [],
                        "last": "Li",
                        "suffix": ""
                    }
                ],
                "year": 2016,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {
                    "arXiv": [
                        "arXiv:1607.05809"
                    ]
                },
                "num": null,
                "urls": [],
                "raw_text": "Kun Xiong, Anqi Cui, Zefeng Zhang, and Ming Li. 2016. Neural contextual conversation learning with labeled question-answering pairs. arXiv preprint arXiv:1607.05809.",
                "links": null
            },
            "BIBREF27": {
                "ref_id": "b27",
                "title": "Dirichlet latent variable hierarchical recurrent encoderdecoder in dialogue generation",
                "authors": [
                    {
                        "first": "Min",
                        "middle": [],
                        "last": "Zeng",
                        "suffix": ""
                    },
                    {
                        "first": "Yisen",
                        "middle": [],
                        "last": "Wang",
                        "suffix": ""
                    },
                    {
                        "first": "Yuan",
                        "middle": [],
                        "last": "Luo",
                        "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 (EMNLP-IJCNLP)",
                "volume": "",
                "issue": "",
                "pages": "1267--1272",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Min Zeng, Yisen Wang, and Yuan Luo. 2019. Dirichlet latent variable hierarchical recurrent encoder- decoder in dialogue generation. In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Pro- cessing (EMNLP-IJCNLP), pages 1267-1272.",
                "links": null
            },
            "BIBREF28": {
                "ref_id": "b28",
                "title": "Hierarchy response learning for neural conversation generation",
                "authors": [
                    {
                        "first": "Bo",
                        "middle": [],
                        "last": "Zhang",
                        "suffix": ""
                    },
                    {
                        "first": "Xiaoming",
                        "middle": [],
                        "last": "Zhang",
                        "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 (EMNLP-IJCNLP)",
                "volume": "",
                "issue": "",
                "pages": "1772--1781",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Bo Zhang and Xiaoming Zhang. 2019. Hierarchy response learning for neural conversation generation. In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP), pages 1772-1781.",
                "links": null
            },
            "BIBREF29": {
                "ref_id": "b29",
                "title": "Learning to control the specificity in neural response generation",
                "authors": [
                    {
                        "first": "Ruqing",
                        "middle": [],
                        "last": "Zhang",
                        "suffix": ""
                    },
                    {
                        "first": "Jiafeng",
                        "middle": [],
                        "last": "Guo",
                        "suffix": ""
                    },
                    {
                        "first": "Yixing",
                        "middle": [],
                        "last": "Fan",
                        "suffix": ""
                    },
                    {
                        "first": "Yanyan",
                        "middle": [],
                        "last": "Lan",
                        "suffix": ""
                    },
                    {
                        "first": "Jun",
                        "middle": [],
                        "last": "Xu",
                        "suffix": ""
                    },
                    {
                        "first": "Xueqi",
                        "middle": [],
                        "last": "Cheng",
                        "suffix": ""
                    }
                ],
                "year": 2018,
                "venue": "Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics",
                "volume": "1",
                "issue": "",
                "pages": "1108--1117",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Ruqing Zhang, Jiafeng Guo, Yixing Fan, Yanyan Lan, Jun Xu, and Xueqi Cheng. 2018. Learning to control the specificity in neural response generation. In Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 1108-1117.",
                "links": null
            },
            "BIBREF30": {
                "ref_id": "b30",
                "title": "Recosa: Detecting the relevant contexts with self-attention for multi-turn dialogue generation",
                "authors": [
                    {
                        "first": "Hainan",
                        "middle": [],
                        "last": "Zhang",
                        "suffix": ""
                    },
                    {
                        "first": "Yanyan",
                        "middle": [],
                        "last": "Lan",
                        "suffix": ""
                    },
                    {
                        "first": "Liang",
                        "middle": [],
                        "last": "Pang",
                        "suffix": ""
                    },
                    {
                        "first": "Jiafeng",
                        "middle": [],
                        "last": "Guo",
                        "suffix": ""
                    },
                    {
                        "first": "Xueqi",
                        "middle": [],
                        "last": "Cheng",
                        "suffix": ""
                    }
                ],
                "year": 2019,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {
                    "arXiv": [
                        "arXiv:1907.05339"
                    ]
                },
                "num": null,
                "urls": [],
                "raw_text": "Hainan Zhang, Yanyan Lan, Liang Pang, Jiafeng Guo, and Xueqi Cheng. 2019. Recosa: Detect- ing the relevant contexts with self-attention for multi-turn dialogue generation. arXiv preprint arXiv:1907.05339.",
                "links": null
            }
        },
        "ref_entries": {
            "TABREF1": {
                "html": null,
                "content": "<table><tr><td>11.73%</td><td>2.32%</td><td/><td/><td/><td/><td>7.49%</td></tr><tr><td/><td/><td/><td/><td/><td/><td>16.60%</td></tr><tr><td/><td/><td/><td/><td/><td/><td>42.52%</td></tr><tr><td/><td colspan=\"2\">85.96%</td><td/><td/><td/></tr><tr><td/><td/><td/><td/><td/><td/><td>33.40%</td></tr><tr><td colspan=\"2\">\u4e2d\u6027 \u79ef\u6781 \u6d88\u6781</td><td/><td/><td/><td/><td>\u9648\u8ff0 \u63d0\u95ee \u6307\u793a \u627f\u8bfa</td></tr><tr><td/><td>(a) \u60c5\u611f\u7c7b\u522b\u5206\u5e03</td><td/><td/><td/><td/><td>(b) \u610f\u56fe\u7c7b\u522b\u5206\u5e03</td></tr><tr><td/><td/><td/><td/><td/><td/><td>\u6821\u56ed\u751f\u6d3b</td></tr><tr><td/><td>1.99% 3.26%</td><td>7.80%</td><td>4.68%</td><td>5.35%</td><td>13.44%</td><td>\u5de5\u4f5c \u5065\u5eb7</td></tr><tr><td/><td>0.67%</td><td/><td/><td/><td>3.17%</td><td>\u65e5\u5e38\u751f\u6d3b \u4eba\u9645\u5173\u7cfb</td></tr><tr><td/><td/><td/><td/><td/><td/><td>\u6587\u5316\u4e0e\u6559\u80b2</td></tr><tr><td/><td>28.74%</td><td/><td/><td/><td/><td>\u653f\u6cbb</td></tr><tr><td/><td/><td/><td/><td/><td>30.92%</td><td>\u6001\u5ea6\u4e0e\u60c5\u611f</td></tr><tr><td/><td/><td/><td/><td/><td/><td>\u65c5\u6e38</td></tr><tr><td/><td/><td/><td/><td/><td/><td>\u91d1\u878d</td></tr><tr><td/><td/><td/><td colspan=\"3\">(c) \u8bdd\u9898\u7c7b\u522b\u5206\u5e03</td></tr><tr><td colspan=\"7\">\u8fd9\u91ccW a s \uff0cB a s \u4e3a\u6a21\u578b\u53c2\u6570\uff0cP a s \u662f\u610f\u56fe\u9884\u6d4b\u6a21\u578b\u7684\u8f93\u51fa\uff0c\u7528\u4e8e\u610f\u56fe\u5206\u7c7b\u3002 \u56fe 2. \u60c5\u611f\u3001\u610f\u56fe\u3001\u8bdd\u9898\u7c7b\u522b\u6982\u7387\u5206\u5e03</td></tr><tr><td colspan=\"7\">\u56fe\u5e38\u5e38\u4f1a\u53d7\u5230\u524d\u8005\u7684\u5f71\u54cd\uff0c\u800c\u524d\u8005\u662f\u5bf9\u8bdd\u7684\u53d1\u8d77\u65b9\uff0c\u5904\u4e8e\u4e3b\u5bfc\u5730\u4f4d\u3002\u4f8b\u5982\uff0c\u63d0\u95ee\u548c\u9648\u8ff0\u5f80\u5f80\u662f</td></tr><tr><td colspan=\"7\">4.1 \u6570 \u6570 \u6570\u636e \u636e \u636e\u96c6 \u96c6 \u96c6 \u540c\u65f6\u53d1\u751f\u7684\uff0c\u56e0\u4e3a\u5f53\u6709\u4eba\u5411\u6211\u4eec\u63d0\u95ee\u65f6\uff0c\u6211\u4eec\u901a\u5e38\u4e0d\u4f1a\u8f6c\u79fb\u8bdd\u9898\uff0c\u800c\u662f\u793c\u8c8c\u5730\u56de\u590d\u522b\u4eba\u7684\u95ee</td></tr><tr><td colspan=\"7\">\u9898\u3002\u53e6\u5916\u6307\u793a\u548c\u627f\u8bfa\u5f80\u5f80\u4e5f\u662f\u540c\u65f6\u53d1\u751f\u7684\uff0c\u5f53\u6709\u4eba\u5411\u6211\u4eec\u63d0\u51fa\u5efa\u8bae\u65f6\uff0c\u6211\u4eec\u4e00\u822c\u4f1a\u5bf9\u5bf9\u65b9\u6240\u63d0 \u672c\u5b9e\u9a8c\u4f7f\u7528DailyDialog\u5bf9\u8bdd\u8bed\u6599 (Li et al., 2017b)\uff0c\u8be5\u8bed\u6599\u6536\u96c6\u4e8e\u82f1\u8bed\u5b66\u4e60\u7f51\u7ad9\u7684\u5bf9\u8bdd \u7684\u5efa\u8bae\u505a\u51fa\u56de\u5e94\u3002\u56fe 4\u663e\u793a\u4e86\u540c\u4e00\u8bf4\u8bdd\u8005\u524d\u4e00\u53e5\u7684\u60c5\u611f\u786e\u5b9a\u7684\u6761\u4ef6\u4e0b\uff0c\u540e\u4e00\u53e5\u7684\u60c5\u611f\u7c7b\u522b\u7684\u6982 \u7ec3\u4e60\u3002\u8be5\u8bed\u6599\u7684\u57fa\u672c\u7edf\u8ba1\u4fe1\u606f\u5982\u8868 2\u6240\u793a\uff0c\u5171\u5305\u542b13118\u4e2a\u591a\u56de\u5408\u5bf9\u8bdd\uff0c\u5e73\u5747\u6bcf\u7ec4\u5bf9\u8bdd\u8f6e\u6570 \u7387\u5206\u5e03\u3002\u4e3a\u4e86\u907f\u514d\u65e0\u60c5\u611f\u8bdd\u8bed\u7684\u5e72\u6270\uff0c\u6211\u4eec\u53ea\u7edf\u8ba1\u4e86\u8bdd\u8bed\u7684\u60c5\u611f\u4e3a\u79ef\u6781\u6216\u8005\u6d88\u6781\u7684\u60c5\u51b5\u3002\u4ece\u56fe \u7ea6\u4e3a8,\u5e73\u5747\u6bcf\u53e5\u5bf9\u8bdd\u7684\u5355\u8bcd\u6570\u7ea6\u4e3a15\uff0c\u5e73\u5747\u6bcf\u7ec4\u5bf9\u8bdd\u7684\u5355\u8bcd\u6570\u7ea6\u4e3a115\u3002\u8be5\u6570\u636e\u96c6\u4e2d\u7684\u5bf9\u8bdd \u4e2d\u6211\u4eec\u770b\u51fa\u76f8\u540c\u60c5\u611f\u540c\u65f6\u51fa\u73b0\u7684\u6982\u7387\u8fdc\u8fdc\u9ad8\u4e8e\u5176\u4ed6\u60c5\u611f\uff0c\u8fd9\u8bf4\u660e\u540c\u4e00\u8bf4\u8bdd\u8005\u7684\u60c5\u611f\u57fa\u8c03\u901a\u5e38\u662f \u53cd\u6620\u4e86\u6211\u4eec\u7684\u65e5\u5e38\u4ea4\u6d41\u65b9\u5f0f\uff0c\u6db5\u76d6\u4e86\u6211\u4eec\u65e5\u5e38\u751f\u6d3b\u7684\u5404\u79cd\u8bdd\u9898\u3002\u8be5\u8bed\u6599\u4e2d\u7684\u6bcf\u53e5\u8bdd\u90fd\u6807\u6ce8 \u4fdd\u6301\u4e0d\u53d8\u7684\u3002\u57fa\u4e8e\u4e0a\u8ff0\u5206\u6790\uff0c\u6211\u4eec\u53ef\u4ee5\u53d1\u73b0\u5bf9\u5bf9\u8bdd\u7684\u60c5\u611f\u548c\u610f\u56fe\u8bc6\u522b\u5728\u56de\u590d\u751f\u6210\u4e2d\u662f\u975e\u5e38\u91cd\u8981 \u4e86\u60c5\u611f\u548c\u610f\u56fe\u7c7b\u522b\uff0c\u8fd9\u4e9b\u6807\u6ce8\u75313\u540d\u8bed\u8a00\u4e13\u5bb6\u5171\u540c\u5b8c\u6210\uff0c\u5177\u6709\u8f83\u9ad8\u7684\u53ef\u9760\u6027\u3002\u5176\u4e2d\u4e3b\u9898\u5206 \u4e3a10\u7c7b\uff1a\u6821\u56ed\u751f\u6d3b(School Life)\u3001\u5de5\u4f5c(Work)\u3001\u5065\u5eb7(Health)\u3001\u65e5\u5e38\u751f\u6d3b(Ordinary Life)\u3001\u4eba\u9645 \u7684\u3002</td></tr><tr><td colspan=\"7\">\u5173\u7cfb(Relationship)\u3001\u6587\u5316\u4e0e\u6559\u80b2(Culture &amp; Education)\u3001\u653f\u6cbb(Politics)\u3001\u6001\u5ea6\u4e0e\u60c5\u611f(Attitude 0.4 0.6 \u8ff0(Inform)\u3001\u8be2\u95ee(Question)\u3001\u6307\u793a(Directive)\u3001\u8bb8\u8bfa(Commissive)\u3002 0.5 0.6 0.8 \u60c5\u611f\uff0c\u5c06\u60c5\u611f\u91cd\u65b0\u5206\u4e3a\u4e2d\u6027(Neutal)\u3001\u79ef\u6781(Positive)\u3001\u6d88\u6781(Negative)\u4e09\u7c7b\uff1b\u610f\u56fe\u5206\u4e3a4\u7c7b\uff1a\u9648 0.7 1 &amp; Emotion)\u3001\u65c5\u6e38(Tourism)\u3001\u91d1\u878d(Finance)\uff1b\u60c5\u611f\u5206\u4e3a7\u7c7b\uff0c\u5728\u672c\u5b9e\u9a8c\u4e2d\u6211\u4eec\u4e3a\u4e86\u66f4\u597d\u7684\u8bc6\u522b 0.8 1.2</td></tr><tr><td>0.1 0.2 0.3</td><td colspan=\"3\">\u5bf9 \u5bf9 \u5bf9\u8bdd \u8bdd \u8bdd\u603b \u603b \u603b\u6570 \u6570 \u6570</td><td/><td/><td>0.2 13118 0.4</td></tr><tr><td colspan=\"7\">0 \u56fe 3. \u540c\u4e00\u8f6e\u5bf9\u8bdd\u4e2d\u610f\u56fe\u7684\u5f71\u54cd \u9648\u8ff0 \u8be2\u95ee \u6307\u793a \u627f\u8bfa \u9648\u8ff0 \u8be2\u95ee \u6307\u793a \u627f\u8bfa \u5e73 \u5e73 \u8868 2. DailyDialog\u57fa\u672c\u4fe1\u606f\u7edf\u8ba1 0 \u56fe 4. \u540c\u4e00\u8bf4\u8bdd\u8005\u60c5\u611f\u7684\u5f71\u54cd \u79ef\u6781 \u6d88\u6781 \u79ef\u6781 \u6d88\u6781 \u672c\u6587\u4f7f\u7528BLEU(Bilingual Evaluation Under-study)</td></tr><tr><td colspan=\"7\">\u672c\u5b9e\u9a8c\u4e2d\uff0c\u6211\u4eec\u7814\u7a76\u56db\u8f6e\u5bf9\u8bdd\uff0c\u56e0\u6b64\u6211\u4eec\u8fc7\u6ee4\u4e86\u5c11\u4e8e\u516b\u53e5\u7684\u5bf9\u8bdd\uff0c\u5e76\u622a\u53d6\u5927\u4e8e\u7b49\u4e8e\u516b\u53e5\u5bf9</td></tr><tr><td colspan=\"7\">\u8bdd\u4e2d\u7684\u524d\u516b\u53e5\u3002\u5728\u4ee5\u4e0a\u6761\u4ef6\u4e0b\uff0c\u6311\u90095835\u7ec4\u5bf9\u8bdd\u4f5c\u4e3a\u8bad\u7ec3\u96c6\uff0c200\u7ec4\u4f5c\u4e3a\u6d4b\u8bd5\u96c6\u3002\u56fe 2\u4e2d\u5206\u522b\u7ed9</td></tr><tr><td colspan=\"7\">\u51fa\u4e86\u8fc7\u6ee4\u540e\u7684\u6570\u636e\u96c6\u4e2d\u8bdd\u9898\u3001\u60c5\u611f\u3001\u610f\u56fe\u7684\u7c7b\u522b\u6982\u7387\u5206\u5e03\u3002\u4ece\u56fe\u4e2d\u6211\u4eec\u53d1\u73b0\uff0c\u5bf9\u8bdd\u4e2d\u60c5\u611f\u4e3a\u4e2d</td></tr><tr><td colspan=\"7\">\u6027\u7684\u8bdd\u8bed\u5360\u5927\u591a\u6570\uff0c\u610f\u56fe\u4e3a\u9648\u8ff0\u548c\u8be2\u95ee\u7684\u8bdd\u8bed\u5360\u6bd4\u9ad8\u4e8e\u5176\u4ed6\u4e24\u79cd\u610f\u56fe\u3002</td></tr><tr><td colspan=\"7\">\u56fe 3\u5c55\u793a\u4e86\u540c\u4e00\u8f6e\u5bf9\u8bdd\u4e2d\u524d\u4e00\u53e5\u7684\u610f\u56fe\u786e\u5b9a\u65f6\uff0c\u540e\u4e00\u53e5\u8bdd\u7684\u610f\u56fe\u7c7b\u522b\u7684\u6982\u7387\u5206\u5e03\u3002\u4ece\u56fe\u4e2d</td></tr><tr><td colspan=\"7\">\u6211\u4eec\u53d1\u73b0\u5bf9\u8bdd\u4e2d\u4e0d\u540c\u7684\u89d2\u8272\u4e4b\u95f4\u7684\u610f\u56fe\u662f\u76f8\u4e92\u5f71\u54cd\u3002\u5bf9\u8bdd\u53d1\u751f\u5728\u524d\u540e\u4e24\u4e2a\u89d2\u8272\u4e4b\u95f4\uff0c\u540e\u8005\u7684\u610f</td></tr></table>",
                "num": null,
                "text": "\u5e73\u5747 \u5747 \u5747\u6bcf \u6bcf \u6bcf\u7ec4 \u7ec4 \u7ec4\u5bf9 \u5bf9 \u5bf9\u8bdd \u8bdd \u8bdd\u8f6e \u8f6e \u8f6e\u6570 \u6570 \u6570 7.9 \u5e73 \u5e73 \u5e73\u5747 \u5747 \u5747\u6bcf \u6bcf \u6bcf\u4e2a \u4e2a \u4e2a\u5b50 \u5b50 \u5b50\u53e5 \u53e5 \u53e5\u7684 \u7684 \u7684\u5355 \u5355 \u5355\u8bcd \u8bcd \u8bcd\u6570 \u6570 \u6570 14.6 \u5e73 \u5e73 \u5e73\u5747 \u5747 \u5747\u6bcf \u6bcf \u6bcf\u7ec4 \u7ec4 \u7ec4\u5bf9 \u5bf9 \u5bf9\u8bdd \u8bdd \u8bdd\u7684 \u7684 \u7684\u5355 \u5355 \u5355\u8bcd \u8bcd \u8bcd\u6570 \u6570 \u6570 114.7",
                "type_str": "table"
            },
            "TABREF2": {
                "html": null,
                "content": "<table><tr><td colspan=\"7\">2) 3) \u6211\u4eec\u7684\u6a21\u578b\u7684\u5b9e\u9a8c\u7ed3\u679c\u8d85\u8fc7\u4e86\u6240\u6709\u7684\u57fa\u51c6\u6a21\u578b\uff0c\u5145\u5206\u8bf4\u660e\u4e86\u6211\u4eec\u63d0\u51fa\u7684\u6a21\u578b\u80fd\u6709\u6548\u5730\u63d0\u9ad8\u751f</td></tr><tr><td colspan=\"2\">\u6210\u7684\u5bf9\u8bdd\u56de\u590d\u7684\u8d28\u91cf\u3002</td><td/><td/><td/><td/><td/></tr><tr><td colspan=\"2\">4.4 \u4e0d \u4e0d \u4e0d\u540c \u540c \u540c\u56e0 \u56e0 \u56e0\u7d20 \u7d20 \u7d20\u7684 \u7684 \u7684\u5f71 \u5f71 \u5f71\u54cd \u54cd \u54cd</td><td/><td/><td/><td/><td/></tr><tr><td colspan=\"7\">\u4e3a\u4e86\u9a8c\u8bc1\u6211\u4eec\u6a21\u578b\u7684\u6709\u6548\u6027\uff0c\u5c06\u6211\u4eec\u7684\u6a21\u578b\u4e0e\u5206\u522b\u5355\u72ec\u8003\u8651\u8bdd\u9898\u3001\u60c5\u611f\u3001\u610f\u56fe\u9884\u6d4b\u6a21\u578b\u8fdb</td></tr><tr><td colspan=\"7\">\u884c\u5bf9\u6bd4\u5b9e\u9a8c\u3002\u5b9e\u9a8c\u7ed3\u679c\u5982\u8868 5\u6240\u793a\u3002\u6211\u4eec\u5171\u6d89\u53ca\u4e865\u7ec4\u5bf9\u6bd4\u5b9e\u9a8c\u3002Multi\u6a21\u578b\u5728\u4e0a\u4e00\u8282\u4e2d\u5df2\u7ecf\u4ecb</td></tr><tr><td colspan=\"7\">\u7ecd\uff0c\u4e0d\u518d\u8d58\u8ff0\uff1bJoint Topic\u6a21\u578b\u662f\u5728Multi\u6a21\u578b\u7684\u57fa\u7840\u4e0a\u52a0\u5165\u5bf9\u8bdd\u610f\u56fe\u7684\u8bc6\u522b\uff1bJoint Senti\u6a21\u578b</td></tr><tr><td colspan=\"7\">\u662f\u5728Multi\u6a21\u578b\u7684\u57fa\u7840\u4e0a\u52a0\u5165\u8bdd\u8bed\u60c5\u611f\u7684\u8bc6\u522b\uff1bJoint Act\u6a21\u578b\u662f\u5728Multi\u6a21\u578b\u7684\u57fa\u7840\u4e0a\u52a0\u5165\u8bdd\u8bed</td></tr><tr><td>\u610f\u56fe\u7684\u8bc6\u522b\u3002</td><td/><td/><td/><td/><td/><td/></tr><tr><td>\u6a21 \u6a21 \u6a21\u578b \u578b \u578b\u540d \u540d \u540d\u79f0 \u79f0 \u79f0</td><td colspan=\"6\">BLEU-1 BLEU-2 BLEU-3 Average Greedy Extrema</td></tr><tr><td>Multi</td><td>15.64</td><td>6.48</td><td>5.33</td><td>62.14</td><td>44.33</td><td>37.20</td></tr><tr><td>Joint Topic</td><td>18.01</td><td>8.87</td><td>7.60</td><td>63.34</td><td>45.32</td><td>39.26</td></tr><tr><td>Joint Senti</td><td>16.89</td><td>7.74</td><td>6.34</td><td>62.74</td><td>43.93</td><td>37.32</td></tr><tr><td>Joint Act</td><td>17.97</td><td>8.77</td><td>7.46</td><td>63.22</td><td>45.50</td><td>38.84</td></tr><tr><td>JTEA(Ours)</td><td>19.41</td><td>10.40</td><td>9.01</td><td>64.56</td><td>48.13</td><td>41.46</td></tr><tr><td/><td/><td colspan=\"2\">\u8868 5. \u4e0d\u540c\u56e0\u7d20\u7684\u5f71\u54cd</td><td/><td/><td/></tr><tr><td colspan=\"7\">\u4ece\u8868\u4e2d\u53ef\u4ee5\u53d1\u73b0\uff0c\u6240\u6709\u5177\u6709\u7ea6\u675f\u6761\u4ef6\u7684\u6a21\u578b(Joint Topic, Joint Senti, Joint Act, JTEA)\u90fd</td></tr><tr><td colspan=\"7\">\u4f18\u4e8e\u57fa\u51c6Multi\u6a21\u578b\uff0c\u8fd9\u8868\u660e\u6240\u6709\u7684\u7ea6\u675f\u6761\u4ef6\u5bf9\u4e8e\u56de\u590d\u751f\u6210\u90fd\u662f\u6709\u6548\u7684\u3002\u53e6\u5916\uff0c\u6211\u4eec\u7684\u6a21\u578b\u5373\u8003</td></tr><tr><td colspan=\"7\">\u8651\u6240\u6709\u7684\u7ea6\u675f\u6761\u4ef6\u4f18\u4e8e\u5355\u72ec\u8003\u8651\u6bcf\u4e2a\u7ea6\u675f\u6761\u4ef6\u7684\u6a21\u578b\uff0c\u8fd9\u8868\u660e\u6211\u4eec\u5e94\u8be5\u96c6\u6210\u6240\u6709\u7ea6\u675f\u6761\u4ef6\u6765\u751f</td></tr><tr><td colspan=\"7\">\u6210\u66f4\u9ad8\u8d28\u91cf\u7684\u56de\u590d\u3002\u6211\u4eec\u8fd8\u53ef\u4ee5\u53d1\u73b0\u60c5\u611f\u7684\u7ea6\u675f\u76f8\u5bf9\u4e8e\u4e3b\u9898\u548c\u610f\u56fe\u7684\u7ea6\u675f\u6548\u679c\u8981\u7a0d\u5dee\u4e00\u70b9\uff0c\u662f</td></tr><tr><td colspan=\"7\">\u56e0\u4e3a\u5728\u6211\u4eec\u7684\u8bed\u6599\u4e2d\u7edd\u5927\u90e8\u5206\u8bdd\u8bed\u7684\u60c5\u611f\u90fd\u662f\u4e2d\u7acb\u7684\uff0c\u60c5\u611f\u5bf9\u56de\u590d\u7684\u5f71\u54cd\u76f8\u5bf9\u8f83\u5c0f\u3002</td></tr><tr><td colspan=\"7\">\u4f7f\u7528\u81ea\u6ce8\u610f\u529b\u673a\u5236\u6765\u66f4\u65b0\u4e0a\u4e0b\u6587\u548c\u88ab\u5c4f\u853d\u7684\u54cd\u5e94\u8868 \u793a\uff0c\u5e76\u5728\u89e3\u7801\u8fc7\u7a0b\u4e2d\u4f7f\u7528\u4e0a\u4e0b\u6587\u548c\u54cd\u5e94\u8868\u793a\u4e4b\u95f4\u7684\u6ce8\u610f\u529b\u6743\u91cd\u3002 4.5 \u6848 \u6848 \u6848\u4f8b \u4f8b \u4f8b\u5206 \u5206 \u5206\u6790 \u6790 \u6790</td></tr><tr><td colspan=\"7\">\u6211\u4eec\u5bf9\u57fa\u7ebfMulti\u6a21\u578b\u548c\u6211\u4eec\u7684\u6a21\u578b\u751f\u6210\u7684\u5bf9\u8bdd\u56de\u590d\u8fdb\u884c\u5bf9\u6bd4\u5206\u6790\uff0c\u8868 6\u7ed9\u51fa\u4e86\u4e09\u7ec4\u5bf9\u8bdd\u793a HRG\u6a21 \u6a21 \u6a21\u578b \u578b \u578b\uff1a \uff1a \uff1a (Zhang and Zhang, 2019)\u91c7\u7528\u5206\u5c42\u54cd\u5e94\u751f\u6210\u6846\u67b6\u4ee5\u81ea\u7136\u548c\u8fde\u8d2f\u7684\u65b9\u5f0f\u6355\u83b7\u5bf9 \u8bdd\u610f\u56fe\u3002 \u4f8b\uff0c\u6211\u4eec\u622a\u53d6\u4e86\u5bf9\u8bdd\u7684\u4e3b\u8981\u5185\u5bb9\u4e14\u5bf9\u8bdd\u5185\u5bb9\u90fd\u5df2\u7ffb\u8bd1\u4e3a\u4e2d\u6587\u5c55\u793a\u3002</td></tr><tr><td>\u6a21 \u6a21 \u6a21\u578b \u578b \u578b\u540d \u540d \u540d\u79f0 \u79f0 \u79f0</td><td colspan=\"6\">BLEU-1 BLEU-2 BLEU-3 Average Greedy Extrema</td></tr><tr><td>Seq2Seq</td><td>12.94</td><td>5.64</td><td>4.80</td><td>57.88</td><td>42.01</td><td>36.06</td></tr><tr><td>Multi</td><td>15.64</td><td>6.48</td><td>5.33</td><td>62.14</td><td>44.33</td><td>37.20</td></tr><tr><td>Dir-VHRED</td><td>10.90</td><td>3.82</td><td>2.26</td><td>59.92</td><td>41.26</td><td>32.70</td></tr><tr><td>ReCoSa</td><td>17.24</td><td>8.37</td><td>6.89</td><td>62.63</td><td>46.59</td><td>40.42</td></tr><tr><td>HRG</td><td>17.58</td><td>7.55</td><td>5.88</td><td>63.29</td><td>45.34</td><td>38.50</td></tr><tr><td>JTEA(Ours)</td><td>19.41</td><td>10.40</td><td>9.01</td><td>64.56</td><td>48.13</td><td>41.46</td></tr><tr><td/><td/><td colspan=\"2\">\u8868 4. \u4e0e\u57fa\u7ebf\u6a21\u578b\u6bd4\u8f83</td><td/><td/><td/></tr><tr><td colspan=\"7\">\u8868 4\u5c55\u793a\u4e86\u6211\u4eec\u7684\u6a21\u578b\u548c\u57fa\u7ebf\u6a21\u578b\u7684\u6bd4\u8f83\u7ed3\u679c\u3002\u4ece\u8868\u4e2d\u6211\u4eec\u53ef\u4ee5\u5f97\u51fa\u4ee5\u4e0b\u7ed3\u8bba\uff1a</td></tr><tr><td colspan=\"7\">1) Multi\u6a21\u578b\u6bd4Seq2Seq\u6a21\u578b\u8868\u73b0\u597d\u3002\u56e0\u4e3aSeq2Seq\u6a21\u578b\u662f\u5c06\u6240\u6709\u5386\u53f2\u53e5\u5b50\u4fe1\u606f\u62fc\u63a5\u4f5c\u4e3a\u8f93\u5165\u5e8f</td></tr><tr><td colspan=\"7\">\u5217\uff0c\u5bfc\u81f4\u524d\u9762\u5b50\u53e5\u7684\u8bed\u4e49\u4fe1\u606f\u88ab\u9010\u6e10\u7a00\u91ca\u6389\uff0c\u751f\u6210\u7684\u4e2d\u95f4\u8bed\u4e49\u5411\u91cf\u4e0d\u80fd\u5145\u5206\u63d0\u53d6\u5386\u53f2\u4fe1\u606f</td></tr><tr><td colspan=\"7\">\u4e2d\u7684\u7279\u5f81\u3002\u540c\u65f6\u4e5f\u8bc1\u660e\u4e86\u5c06\u5386\u53f2\u4fe1\u606f\u4e2d\u7684\u6bcf\u4e00\u4e2a\u5b50\u53e5\u5206\u522b\u7ecf\u8fc7LSTM\u7f51\u7edc\uff0c\u53d6\u5e73\u5747\u503c\u4f5c\u4e3a</td></tr><tr><td colspan=\"2\">\u4e2d\u95f4\u8bed\u4e49\u5411\u91cf\u66f4\u52a0\u5408\u7406\u3002</td><td/><td/><td/><td/><td/></tr></table>",
                "num": null,
                "text": "\u6211 \u4eec \u7684 \u6a21 \u578b \u76f8 \u8f83 \u4e8eMulti\u6a21 \u578b \uff0cBLEU-1\u3001BLEU-2\u3001BLEU-3\u503c \u5206 \u522b \u63d0 \u5347 \u4e863.77\u30014.76\u30013.68\u4e2a \u767e \u5206 \u70b9 \uff0cAverage\u3001Greedy\u3001Extrema\u5206 \u522b \u63d0 \u5347 \u4e862.42\u30013.8\u30014.26\u4e2a \u767e\u5206\u70b9\u3002\u800c\u6211\u4eec\u7684\u6a21\u578b\u662f\u5728Multi\u6a21\u578b\u7684\u57fa\u7840\u4e0a\u4e86\u5f15\u5165\u4e86\u4e3b\u9898\u3001\u60c5\u611f\u548c\u610f\u56fe\u7684\u8bc6\u522b\uff0c\u8fd9\u8bc1\u660e \u4e86\u5bf9\u5bf9\u8bdd\u56de\u590d\u8fdb\u884c\u7ea6\u675f\u53ef\u4ee5\u6709\u6548\u5730\u63d0\u9ad8\u751f\u6210\u56de\u590d\u7684\u8d28\u91cf\u3002",
                "type_str": "table"
            }
        }
    }
}