{
    "paper_id": "O06-1012",
    "header": {
        "generated_with": "S2ORC 1.0.0",
        "date_generated": "2023-01-19T08:07:41.736280Z"
    },
    "title": "Disfluency Correction of Spontaneous Speech using Conditional Random Fields with Variable Length Features",
    "authors": [
        {
            "first": "",
            "middle": [],
            "last": "\uf96e\u745e\u5cf0",
            "suffix": "",
            "affiliation": {
                "laboratory": "",
                "institution": "Far East University",
                "location": {
                    "settlement": "Tainan County",
                    "country": "Taiwan"
                }
            },
            "email": ""
        },
        {
            "first": "",
            "middle": [],
            "last": "\u3001\u5433\u5b97\u61b2",
            "suffix": "",
            "affiliation": {
                "laboratory": "",
                "institution": "National Cheng-Kung University",
                "location": {
                    "settlement": "Tainan",
                    "country": "Taiwan"
                }
            },
            "email": ""
        },
        {
            "first": "",
            "middle": [],
            "last": "\u3001\u5433\u7dad\u5f65",
            "suffix": "",
            "affiliation": {
                "laboratory": "",
                "institution": "National Cheng-Kung University",
                "location": {
                    "settlement": "Tainan",
                    "country": "Taiwan"
                }
            },
            "email": ""
        }
    ],
    "year": "",
    "venue": null,
    "identifiers": {},
    "abstract": "",
    "pdf_parse": {
        "paper_id": "O06-1012",
        "_pdf_hash": "",
        "abstract": [],
        "body_text": [
            {
                "text": "= \u239b \u239e = \u239c \u239f \u239c \u239f \u239d \u23a0 \u239b \u239e \u2245 \u239c \u239f \u239c \u239f \u239d \u23a0 \u2211 \u2211 (1) 3. \uf967\u5b9a\u9577\ufa01\u7279\u5fb5\u4e4b\u689d\u4ef6\u96a8\u6a5f\u57df \u689d\u4ef6\u96a8\u6a5f\u57df \u689d\u4ef6\u96a8\u6a5f\u57df\u70ba\u4e00\u7a2e\u7121\u5411\u5716(Undirected Graphical)\u7684\u6a21\u578b\uff0c\u53ef\u88ab\u7528\uf92d\u4f30\u7b97\u7d66\u4e88\u4e00\u89c0\u6e2c\u5e8f\uf99c\uff0c\u5f97\u5230\u76f8 \u5c0d\u61c9\u7684\uf9fa\u614b\u5e8f\uf99c\u5176\u4ea4\u96c6\u7684\u6a5f\uf961\u5206\u4f48\u3002\u5176\u6982\uf9a3\u662f\u4ee5\u96a8\u6a5f\u57df\u70ba\u57fa\u790e\uff0c\u52a0\u4e0a\u5168\u57df\u88ab\u9650\u5236\u65bc X \u9019\u500b\u689d\u4ef6\uff0c \u7a31\u4e4b\u4e00\u689d\u4ef6\u96a8\u6a5f\u57df\uff0cX \u70ba\u89c0\u6e2c\u5e8f\uf99c\u3002\u6b63\u5f0f\u7684\uf92d\uf96f\uff0c\u6211\u5011\u5b9a\u7fa9\u4e00\u5716 G=(S,E)\u70ba\u4e00\u7121\u5411\u5716\uff0cS \u70ba\u6240\u6709 \u9ede\u4e4b\u96c6\u5408\uff0c\u6bcf\u500b\u9ede\u7686\u70ba\u96a8\u6a5f\u8b8a\uf969\uff0c\u6211\u5011\u53ef\u5c07\u67d0\u500b\u9ede Sv \u770b\u6210\uf9fa\u614b\u5e8f\uf99c Y \u4e0a\u7684\u67d0\u500b\uf9fa\u614b Yv\u3002\uf974\u6bcf \u500b\u96a8\u6a5f\u8b8a\uf969 Yv \ufa26\u9075\u5b88\u99ac\u53ef\u592b\u539f\u5247\uff0c\u4e5f\u5c31\u662f\uf96f\u7d66\u4e88 X \u548c\u6240\u6709\u5176\u4ed6\u96a8\u6a5f\u8b8a\uf969 Y{u|u\u2260v}\u7684\u689d\u4ef6\u4e4b\u4e0b\uff0c \u5f97\u5230\u96a8\u6a5f\u8b8a\uf969 Yv \u7684\u6a5f\uf961 { } ( ) V Y X, Y u v \u2208 \u2260 v , u , v u , | p (2) \u76f8\u7b49\u65bc\u7d66\u4e88 X \u548c Yv \u7684\u9130\u5c45\u9ede\u7684\u689d\u4ef6\u4e4b\u4e0b\uff0c\u5f97\u5230\u96a8\u6a5f\u8b8a\uf969 Yv \u7684\u6a5f\uf961\uff0c { } ( ) V Y X, Y u v \u2208 = v , u ), v ( neighbor u , | p (3) \u5247(X,Y)\u70ba\u4e00\u689d\u4ef6\u96a8\u6a5f\u57df\u3002 \uf9e4\uf941\u4e0a\uf92d\uf96f\uff0c\u5716 G \u7684\u7d50\u69cb\u53ef\u4ee5\u662f\u4efb\u610f\u7684\uff0c\u7136\u800c\uff0c\u7576\u7528\uf92d\u5c0d\u5e8f\uf99c\u5efa\u6a21\u6642\uff0c\u6700\u7c21\u55ae\u4e14\u6700\u666e\u901a\u7684 \u5716\u7684\u7d50\u69cb\u5247\u70ba\u5f62\u6210\u4e00\u500b\u7c21\u55ae\u7684\u4e00\u968e\u93c8(First-Order chain )\uff0c\u5982\u4e0b\u5716\u6240\u793a\uff0c\u5176\u4e2d W \u70ba\u89c0\u6e2c\u503c\uff0cS \u70ba \uf9fa\u614b\u5e8f\uf99c\u3002 M M M M M \u5716\u4e09\uff0c\u689d\u4ef6\u96a8\u6a5f\u57df\u793a\u610f\u5716 \u65bc\u662f\u5728\u7d66\u4e88\u89c0\u6e2c\u5e8f\uf99c X\uff0c\u5f97\u5230\u5c0d\u61c9\uf9fa\u614b\u5e8f\uf99c S \u7684\u6a5f\uf961\u70ba: ( ) ( ) ( ) ( ) ( ) ( ) W W t 1 t t k k k k t k t k 1 P S |W exp f s ,s , g s , Z \u03bb \u03bc \u2212 \u239b \u239e = + \u239c \u239f \u239d \u23a0 \u2211\u2211 \u2211\u2211 (4) \u5176\u4e2d ( ) ( ) ( ) W 1 , s , s f t t k \u2212 \u70ba\u6574\u500b\u89c0\u6e2c\u5e8f\uf99c\u548c\u5728\uf9fa\u614b\u5e8f\uf99c\u4e2d\u4f4d\u7f6e t-1 \u7684\uf9fa\u614b\u8f49\u5230\u4f4d\u7f6e t \u7684\uf9fa\u614b\u8f49\u79fb\u7279 \u5fb5\u51fd\uf969\u5b9a\u7fa9\u70ba\uff1b ( ) ( ) ( ) ( ) ( ) \u23a9 \u23a8 \u23a7 = \u2227 = = \u2212 \u2212 otherwise ' s s s s if , s , s f t t t t k 0 1 W 1 1 (5) \u800c ( ) ( ) W , s g t k \u70ba\uf9fa\u614b\u5e8f\uf99c\u4e2d\u4f4d\u7f6e t \u548c\u89c0\u6e2c\u5e8f\uf99c\u7684\uf9fa\u614b\u7279\u5fb5\u51fd\uf969: ( ) ( ) ( ) ( ) \u23a9 \u23a8 \u23a7 = \u2227 = = otherwise w s s if , s g t t k 0 W 1 W t (6) k \u03bb \u4ee5\u53ca k \u03bc \u70ba\u5f9e\u8a13\uf996\u8a9e\uf9be\u4e2d\u4f30\u51fa\uf92d\u7684\uf96b\uf969\uff0cZ \u662f\u6b63\u898f\u5316\u4fc2\uf969\uff0c\u70ba\u6240\u6709\u53ef\u80fd\u7684\uf9fa\u614b\u5e8f\uf99c\u7684\u6a5f\uf961\u7e3d \u548c\uff0c\u5176\u5b9a\u7fa9\u5982\u4e0b\u5f0f\u6240\u793a: ( ) ( ) ( ) ( ) ( ) \u2211 \u2211 \u2211 \u2211 \u2211 \u239f \u23a0 \u239e \u239c \u239d \u239b + = \u2212 s , W t k t k k t k t t k k , s g , s , s f exp W W Z 1 \u03bc \u03bb (7) \u5c31\u4fee\u6b63\uf967\uf9ca\u66a2\u8a9e\uf9ca\u800c\u8a00\uff0c\u6211\u5011\u53ef\u4ee5\u628a\u89c0\u6e2c\u5e8f\uf99c X \u7576\u4f5c\u662f\u4ee5\u8a5e\u69cb\u6210\u7684\u5e8f\uf99c\uff0c\uf9fa\u614b\u5e8f\uf99c Y \u70ba\u4ee5 1 \u548c 0 \u7d44\u6210\u4e4b\u5e8f\uf99c\uff0c\uf974\u5728\u89c0\u6e2c\u5e8f\uf99c\u4e2d\u4f4d\u7f6e t \u7684\u8a5e\u6240\u5c0d\u61c9\u7684\uf9fa\u614b\u70ba",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "",
                "sec_num": null
            },
            {
                "text": "( ) ( ) ( ) ( ) ( ) ( ) W W t 1 t t k k p p k k p t k p,q t k c 1 P S |W exp f s ,s , g s , Z \u03bb \u03bc \u2212 \u239b \u239e \u239c \u239f = + \u239c \u239f \u239d \u23a0 \u2211\u2211\u2211 \u2211\u2211\u2211 (8) \u5176\u4e2d p,q \u70ba\u5c64\u6b21\u500b\uf969\uff0c\u5728\u6b64\u5171\u6709\u8a5e\u3001\u5b57\u5143\uf905\u96c6\u4ee5\u53ca\uf906\u5b50\u9019\u4e09\u7a2e\u5c64\u6b21\uff0c\u800c\u6bcf\u4e00\u5c64\u6b21\u5305\u542b\uf978\u7a2e\uf9fa\u614b\uff0c \u4e00\u7a2e\u70ba 0\uff0c\u4e5f\u5c31\u662f\u6b64\u5c64\u6b21\u70ba\u53ef\u522a\u9664\u5340\u57df\uff0c\u4fee\u6b63\u6642\u9700\u5c07\u5176\u522a\u9664\uff1b\u53e6\u4e00\u7a2e\u5247\u662f 1\uff0c\u5373\u70ba\uf9ca\u66a2\u90e8\u5206\uff0c\u4fee \u6b63\u6642\u5c07\u5176\u4fdd\uf9cd\u3002\u5176\u4e2d ( ) ( ) ( ) W t 1 t k p p f s ,s , \u2212 \u70ba\u6574\u500b\u89c0\u6e2c\u5e8f\uf99c\u548c\u5728 p \u5c64\u6b21\uf9fa\u614b\u5e8f\uf99c\u4e2d\u4f4d\u7f6e t-1 \u7684\uf9fa\u614b\u8f49 \u5230\u4f4d\u7f6e t \u7684\uf9fa\u614b\u8f49\u79fb\u7279\u5fb5\u51fd\uf969\uff0c\u800c ( ) ( ) W t k p g s , \u70ba p \u5c64\u6b21\uf9fa\u614b\u5e8f\uf99c\u4e2d\u4f4d\u7f6e t",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "",
                "sec_num": null
            },
            {
                "text": "EQUATION",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [
                    {
                        "start": 0,
                        "end": 8,
                        "text": "EQUATION",
                        "ref_id": "EQREF",
                        "raw_str": "( ) ( ) ( ) ( ) ( ) ( ) W W t 1 t t k k p q k k p t k p,q t k p 1 p s |W , exp f s ,s , g s , Z \u03bb \u03bc \u2212 \u239b \u239e \u239c \u239f \u0398 = + \u239c \u239f \u239d \u23a0 \u2211\u2211\u2211 \u2211\u2211\u2211 (14) \u5f9e \u8a13 \uf996 \u8cc7 \uf9be \u7684 \u96c6 \u5408 \u4e2d \uf92d \u4f30 \u7b97 \u51fa \u4f7f \u5f97 \u8a13 \uf996 \u8cc7 \uf9be \u7684 log-\u4f3c \u7136 \ufa01 \u6700 \u5927 \u7684 \u4e00 \u7d44 \uf96b \uf969 ( ) 1 2 1 2 \u03bb ,\u03bb , ;\u03bc ,\u03bc , \u0398 = L L \u3002\u65bc\u662f\u6211\u5011\u5c07\u689d\u4ef6\u96a8\u6a5f\u57df\u7684 ( ) | , p s W \u0398 \u4ee3\u5165 log-\u4f3c\u7136\ufa01\u51fd\uf969\u7684\u5b9a\u7fa9\u5f0f: ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) W,s L log p s|W, W W p W ,s W ,s t 1 t t k k p q k k p W ,s t k p,q t k p p W ,s log p s |W , 1 p W ,s log exp f s ,s , g s , Z \u03bb \u03bc \u2212 \u0398 = \u0398 = \u0398 \u239b \u239e \u239b \u239e \u239c \u239f \u239c \u239f = + \u239c \u239f \u239c \u239f \u239d \u23a0 \u239d \u23a0 \u220f \u2211 \u2211 \u2211\u2211\u2211 \u2211\u2211\u2211 % % % (15) \u7d93\u904e\u6574\uf9e4\u4e4b\u5f8c\u5f97\u5230\u4e0b\u5f0f ( ) ( ) ( ) ( ) ( ) ( ) ( ) W W t 1 t t k k p q k k p W,s t k p,q t k p W p W,s f s ,s , g s , p W log Z \u03bb \u03bc \u2212 \u23a1 \u23a4 \u23a2 \u23a5 + \u2212 \u23a2 \u23a5 \u23a3 \u23a6 \u2211 \u2211\u2211\u2211 \u2211\u2211\u2211 \u2211 % % (16) \u4e4b\u5f8c\u6211\u5011\u5c0d log-\u4f3c\u7136\ufa01\u51fd\uf969\u504f\u5fae\uf96b\uf969 k \u03bb :",
                        "eq_num": "( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) W"
                    }
                ],
                "section": "",
                "sec_num": null
            },
            {
                "text": "EQUATION",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [
                    {
                        "start": 0,
                        "end": 8,
                        "text": "EQUATION",
                        "ref_id": "EQREF",
                        "raw_str": "\u03b2 + = \u23a7 = \u23a8 \u23a9 (31) \u5176\u905e\u8ff4\u95dc\u4fc2\u70ba: ( ) ( ) ( ) i t1 t W W M W \u03b1 \u03b1 \u2212 = (32) \u548c ( ) ( ) ( ) i t1 t1 W M W W \u03b2 \u03b2 + + = (33) \u8207\u7279\u5fb5\u51fd\uf969 k f \u76f8\u4f3c\uff0c\u7279\u5fb5\u51fd\uf969 k g \u5176\u9810\u4f30\u6e2c\u5206\u4f48\u7684\u671f\u671b\u503c\u70ba:",
                        "eq_num": "( ) ( ) ( ) ( ) ( ) ( ) P"
                    }
                ],
                "section": "",
                "sec_num": null
            }
        ],
        "back_matter": [],
        "bib_entries": {
            "BIBREF0": {
                "ref_id": "b0",
                "title": "AutomaticRecognition of Spontaneous Speech for Access to Multilingual Oral HistoryArchives",
                "authors": [
                    {
                        "first": "W",
                        "middle": [],
                        "last": "Byrne",
                        "suffix": ""
                    },
                    {
                        "first": "D",
                        "middle": [],
                        "last": "Doermann",
                        "suffix": ""
                    },
                    {
                        "first": "M",
                        "middle": [],
                        "last": "Franz",
                        "suffix": ""
                    },
                    {
                        "first": "S",
                        "middle": [],
                        "last": "Gustman",
                        "suffix": ""
                    },
                    {
                        "first": "J",
                        "middle": [],
                        "last": "Hajic",
                        "suffix": ""
                    },
                    {
                        "first": "D",
                        "middle": [],
                        "last": "Oard",
                        "suffix": ""
                    },
                    {
                        "first": "M",
                        "middle": [],
                        "last": "Picheny",
                        "suffix": ""
                    },
                    {
                        "first": "J",
                        "middle": [],
                        "last": "Psutka",
                        "suffix": ""
                    },
                    {
                        "first": "B",
                        "middle": [],
                        "last": "Ramabhadran",
                        "suffix": ""
                    },
                    {
                        "first": "D",
                        "middle": [],
                        "last": "Soergel",
                        "suffix": ""
                    },
                    {
                        "first": "T",
                        "middle": [],
                        "last": "Ward",
                        "suffix": ""
                    },
                    {
                        "first": "Z",
                        "middle": [],
                        "last": "Wei-Jin",
                        "suffix": ""
                    }
                ],
                "year": 2004,
                "venue": "IEEE Trans. on Speech and Audio Processing",
                "volume": "12",
                "issue": "4",
                "pages": "420--435",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Byrne, W., D. Doermann, M. Franz, S. Gustman, J. Hajic, D. Oard, M. Picheny, J.Psutka, B. Ramabhadran, D. Soergel, T. Ward, and Z. Wei-Jin, \"AutomaticRecognition of Spontaneous Speech for Access to Multilingual Oral HistoryArchives,\" IEEE Trans. on Speech and Audio Processing, Vol. 12, No. 4, pp.420-435, 2004.",
                "links": null
            },
            "BIBREF1": {
                "ref_id": "b1",
                "title": "Parsing Conversational Speech Using Enhanced Segmentation",
                "authors": [
                    {
                        "first": "J",
                        "middle": [
                            "G"
                        ],
                        "last": "Kahn",
                        "suffix": ""
                    },
                    {
                        "first": "M",
                        "middle": [],
                        "last": "Ostendorf",
                        "suffix": ""
                    },
                    {
                        "first": "C",
                        "middle": [],
                        "last": "Chelba",
                        "suffix": ""
                    }
                ],
                "year": 2004,
                "venue": "Proc. HLT-NAACL",
                "volume": "",
                "issue": "",
                "pages": "125--128",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Kahn, J.G., M. Ostendorf and C. Chelba\" Parsing Conversational Speech Using Enhanced Segmentation.\" Proc. HLT-NAACL, 2004. pp. 125-128.",
                "links": null
            },
            "BIBREF3": {
                "ref_id": "b3",
                "title": "Conversational Telephony System for Rich Transcription",
                "authors": [],
                "year": 2005,
                "venue": "Proc. of IEEE International Conference on Acoustics, Speech, and Signal Processing",
                "volume": "",
                "issue": "",
                "pages": "205--208",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Conversational Telephony System for Rich Transcription.\" In Proc. of IEEE International Conference on Acoustics, Speech, and Signal Processing, 2005 (ICASSP '05). (2005), 205-208.",
                "links": null
            },
            "BIBREF4": {
                "ref_id": "b4",
                "title": "Automatic detection of sentence boundaries and disfluencies based on recognized words",
                "authors": [
                    {
                        "first": "A",
                        "middle": [],
                        "last": "Stolcke",
                        "suffix": ""
                    },
                    {
                        "first": "E",
                        "middle": [],
                        "last": "Shriberg",
                        "suffix": ""
                    },
                    {
                        "first": "R",
                        "middle": [],
                        "last": "Bates",
                        "suffix": ""
                    },
                    {
                        "first": "M",
                        "middle": [],
                        "last": "Ostendorf",
                        "suffix": ""
                    },
                    {
                        "first": "D",
                        "middle": [],
                        "last": "Hakkani",
                        "suffix": ""
                    },
                    {
                        "first": "M",
                        "middle": [],
                        "last": "Plauche",
                        "suffix": ""
                    },
                    {
                        "first": "G",
                        "middle": [],
                        "last": "Tur",
                        "suffix": ""
                    },
                    {
                        "first": "Y",
                        "middle": [],
                        "last": "Lu",
                        "suffix": ""
                    }
                ],
                "year": 1998,
                "venue": "Proc. International Conference on Spoken Language Processing",
                "volume": "",
                "issue": "",
                "pages": "2247--2250",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Stolcke, A., E. Shriberg, R. Bates, M. Ostendorf, D. Hakkani, M. Plauche, G. Tur, and Y. Lu.\" Automatic detection of sentence boundaries and disfluencies based on recognized words,\" In Proc. International Conference on Spoken Language Processing, pages 2247--2250, 1998.",
                "links": null
            },
            "BIBREF5": {
                "ref_id": "b5",
                "title": "Automatic disfluency identification in conversational speech using multiple knowledge sources",
                "authors": [
                    {
                        "first": "Y",
                        "middle": [],
                        "last": "Liu",
                        "suffix": ""
                    },
                    {
                        "first": "E",
                        "middle": [],
                        "last": "Shriberg",
                        "suffix": ""
                    },
                    {
                        "first": "A",
                        "middle": [],
                        "last": "Stolcke",
                        "suffix": ""
                    }
                ],
                "year": 2003,
                "venue": "Proc. Eurospeech",
                "volume": "1",
                "issue": "",
                "pages": "957--960",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Liu, Y., E. Shriberg, and A. Stolcke. \"Automatic disfluency identification in conversational speech using multiple knowledge sources,\" In Proc. Eurospeech, volume 1, pages 957-960, 2003.",
                "links": null
            },
            "BIBREF6": {
                "ref_id": "b6",
                "title": "Statistical language modeling for speech disfluencies",
                "authors": [
                    {
                        "first": "A",
                        "middle": [],
                        "last": "Stolcke",
                        "suffix": ""
                    },
                    {
                        "first": "E",
                        "middle": [],
                        "last": "Shriberg",
                        "suffix": ""
                    }
                ],
                "year": 1996,
                "venue": "Proceedings of the International Conference of Acoustics, Speech, and Signal Processing",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Stolcke, A. and E. Shriberg. \"Statistical language modeling for speech disfluencies\". In Proceedings of the International Conference of Acoustics, Speech, and Signal Processing, 1996.",
                "links": null
            },
            "BIBREF7": {
                "ref_id": "b7",
                "title": "Comparing HMM, Maximum Entropy, and Conditional Random Fields for Disfluency Detection",
                "authors": [
                    {
                        "first": "Y",
                        "middle": [],
                        "last": "Liu",
                        "suffix": ""
                    },
                    {
                        "first": "E",
                        "middle": [],
                        "last": "Shriberg",
                        "suffix": ""
                    },
                    {
                        "first": "A",
                        "middle": [],
                        "last": "Stolcke",
                        "suffix": ""
                    },
                    {
                        "first": "M",
                        "middle": [],
                        "last": "Harper",
                        "suffix": ""
                    }
                ],
                "year": 2005,
                "venue": "Eurospeech",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Liu, Y., E. Shriberg, A. Stolcke, M. Harper\"Comparing HMM, Maximum Entropy, and Conditional Random Fields for Disfluency Detection.\" Eurospeech 2005.",
                "links": null
            },
            "BIBREF8": {
                "ref_id": "b8",
                "title": "Integrating multiple knowledge sources for detecting and correction of repairs in human computer dialog",
                "authors": [
                    {
                        "first": "J",
                        "middle": [],
                        "last": "Bear",
                        "suffix": ""
                    },
                    {
                        "first": "J",
                        "middle": [],
                        "last": "Dowding",
                        "suffix": ""
                    },
                    {
                        "first": "E",
                        "middle": [],
                        "last": "Shriberg",
                        "suffix": ""
                    }
                ],
                "year": 1992,
                "venue": "Proc. of ACL",
                "volume": "",
                "issue": "",
                "pages": "56--63",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Bear, J., J. Dowding, and E. Shriberg, \"Integrating multiple knowledge sources for detecting and correction of repairs in human computer dialog,\" in Proc. of ACL, 1992, pp. 56-63.",
                "links": null
            },
            "BIBREF9": {
                "ref_id": "b9",
                "title": "An efficient repair procedure for quick transcriptions",
                "authors": [
                    {
                        "first": "A",
                        "middle": [],
                        "last": "Stolcke",
                        "suffix": ""
                    },
                    {
                        "first": "W",
                        "middle": [],
                        "last": "Wang",
                        "suffix": ""
                    },
                    {
                        "first": "D",
                        "middle": [],
                        "last": "Vergyri",
                        "suffix": ""
                    },
                    {
                        "first": "V",
                        "middle": [
                            "R R"
                        ],
                        "last": "Gadde",
                        "suffix": ""
                    },
                    {
                        "first": "J",
                        "middle": [],
                        "last": "Zheng",
                        "suffix": ""
                    }
                ],
                "year": 2004,
                "venue": "Proc. Intl. Conf. Spoken Language Processing",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Stolcke, A., W. Wang, D. Vergyri, V. R. R. Gadde, and J. Zheng, \"An efficient repair procedure for quick transcriptions,\" in Proc. Intl. Conf. Spoken Language Processing, (Jeju, Korea), October 2004.",
                "links": null
            },
            "BIBREF10": {
                "ref_id": "b10",
                "title": "Correction of disfluencies in spontaneous speech using a noisy-channel approach",
                "authors": [
                    {
                        "first": "M",
                        "middle": [],
                        "last": "Honal",
                        "suffix": ""
                    },
                    {
                        "first": "T",
                        "middle": [],
                        "last": "Schultz",
                        "suffix": ""
                    }
                ],
                "year": null,
                "venue": "EUROSPEECH-2003",
                "volume": "",
                "issue": "",
                "pages": "2781--2784",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Honal, M., and T. Schultz, , \"Correction of disfluencies in spontaneous speech using a noisy-channel approach,\" In EUROSPEECH-2003, 2781-2784.",
                "links": null
            },
            "BIBREF11": {
                "ref_id": "b11",
                "title": "Edit detection and parsing for transcribed speech",
                "authors": [
                    {
                        "first": "E",
                        "middle": [],
                        "last": "Charniak",
                        "suffix": ""
                    },
                    {
                        "first": "M",
                        "middle": [],
                        "last": "Johnson",
                        "suffix": ""
                    }
                ],
                "year": 2001,
                "venue": "Proceedings of the North American Chapter of the Association for Computational Linguistics annual meeting",
                "volume": "",
                "issue": "",
                "pages": "118--126",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Charniak, E. and M. Johnson. \"Edit detection and parsing for transcribed speech,\" In Proceedings of the North American Chapter of the Association for Computational Linguistics annual meeting, pages 118--126, 2001.",
                "links": null
            },
            "BIBREF12": {
                "ref_id": "b12",
                "title": "A corpus-based study of repair cues in spontaneous speech",
                "authors": [
                    {
                        "first": "C",
                        "middle": [],
                        "last": "Nakatani",
                        "suffix": ""
                    },
                    {
                        "first": "J",
                        "middle": [],
                        "last": "Hirschberg",
                        "suffix": ""
                    }
                ],
                "year": 1994,
                "venue": "Journal of the Acoustical Society of America",
                "volume": "",
                "issue": "",
                "pages": "1603--1616",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Nakatani, C. and J. Hirschberg. \"A corpus-based study of repair cues in spontaneous speech.\" Journal of the Acoustical Society of America, pages 1603--1616, 1994.",
                "links": null
            },
            "BIBREF13": {
                "ref_id": "b13",
                "title": "A lexically-driven algorithm for disfluency detection",
                "authors": [
                    {
                        "first": "M",
                        "middle": [],
                        "last": "Snover",
                        "suffix": ""
                    },
                    {
                        "first": "B",
                        "middle": [],
                        "last": "Dorr",
                        "suffix": ""
                    },
                    {
                        "first": "R",
                        "middle": [],
                        "last": "Schwartz",
                        "suffix": ""
                    }
                ],
                "year": 2004,
                "venue": "Proceedings of Human Language Technology Conference / North American Chapter of the Association for Computational Linguistics annual meeting",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Snover, M., B. Dorr, and R. Schwartz. \"A lexically-driven algorithm for disfluency detection\". In Proceedings of Human Language Technology Conference / North American Chapter of the Association for Computational Linguistics annual meeting, 2004.",
                "links": null
            },
            "BIBREF14": {
                "ref_id": "b14",
                "title": "Detecting structural metadata with decision trees and transformation-based learning",
                "authors": [
                    {
                        "first": "J",
                        "middle": [],
                        "last": "Kim",
                        "suffix": ""
                    },
                    {
                        "first": "S",
                        "middle": [
                            "E"
                        ],
                        "last": "Schwarm",
                        "suffix": ""
                    },
                    {
                        "first": "M",
                        "middle": [],
                        "last": "Ostendorf",
                        "suffix": ""
                    }
                ],
                "year": 2004,
                "venue": "Proceedings of HLT/NAACL",
                "volume": "",
                "issue": "",
                "pages": "137--144",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Kim, J., S. E. Schwarm, and M. Ostendorf,\" Detecting structural metadata with decision trees and transformation-based learning.\" Proceedings of HLT/NAACL 2004, (2004), 137-144.",
                "links": null
            },
            "BIBREF15": {
                "ref_id": "b15",
                "title": "Toward the realization of spontaneous speech recognition -Introduction of a Japanese priority program and preliminary results",
                "authors": [
                    {
                        "first": "S",
                        "middle": [],
                        "last": "Furui",
                        "suffix": ""
                    },
                    {
                        "first": "K",
                        "middle": [],
                        "last": "Maekawa",
                        "suffix": ""
                    },
                    {
                        "first": "H",
                        "middle": [],
                        "last": "Isahara",
                        "suffix": ""
                    },
                    {
                        "first": "T",
                        "middle": [],
                        "last": "Shinozaki",
                        "suffix": ""
                    },
                    {
                        "first": "T",
                        "middle": [],
                        "last": "Ohdaira",
                        "suffix": ""
                    }
                ],
                "year": null,
                "venue": "Proc. ICSLP2000",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Furui, S., K. Maekawa, H. Isahara, T. Shinozaki and T. Ohdaira\"Toward the realization of spontaneous speech recognition -Introduction of a Japanese priority program and preliminary results -\", Proc. ICSLP2000, Beijing.",
                "links": null
            },
            "BIBREF16": {
                "ref_id": "b16",
                "title": "Repairs and Repetitions in Spontaneous Mandarin",
                "authors": [
                    {
                        "first": "S.-C",
                        "middle": [],
                        "last": "Tseng",
                        "suffix": ""
                    }
                ],
                "year": null,
                "venue": "Proceedings of Workshop on Disfluency in Spontaneous Speech (DISS 03). Ed. Robert Eklund. Gothenburg Papers in Theoretical Linguistics 90",
                "volume": "",
                "issue": "",
                "pages": "71--74",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Tseng, S.-C, \" Repairs and Repetitions in Spontaneous Mandarin,\" In Proceedings of Workshop on Disfluency in Spontaneous Speech (DISS 03). Ed. Robert Eklund. Gothenburg Papers in Theoretical Linguistics 90. Pp. 71-74. University of Gothenburg.",
                "links": null
            },
            "BIBREF17": {
                "ref_id": "b17",
                "title": "Important and new features with analysis for disfluency interruption point (IP) detection in spontaneous Mandarin speech",
                "authors": [
                    {
                        "first": "Che-Kuang",
                        "middle": [],
                        "last": "Lin",
                        "suffix": ""
                    },
                    {
                        "first": "",
                        "middle": [],
                        "last": "Tseng",
                        "suffix": ""
                    },
                    {
                        "first": "",
                        "middle": [],
                        "last": "Shu-Chuan",
                        "suffix": ""
                    },
                    {
                        "first": "",
                        "middle": [],
                        "last": "Lee",
                        "suffix": ""
                    },
                    {
                        "first": "",
                        "middle": [],
                        "last": "Lin-Shan",
                        "suffix": ""
                    }
                ],
                "year": null,
                "venue": "DiSS-2005",
                "volume": "",
                "issue": "",
                "pages": "117--121",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Lin, Che-Kuang , Tseng, Shu-Chuan , Lee, Lin-Shan, \"Important and new features with analysis for disfluency interruption point (IP) detection in spontaneous Mandarin speech\", In DiSS-2005, 117-121.",
                "links": null
            },
            "BIBREF19": {
                "ref_id": "b19",
                "title": "Speech act modeling and verification of spontaneous speech with disfluency in a spoken dialogue system",
                "authors": [
                    {
                        "first": "Gwo-Lang",
                        "middle": [],
                        "last": "Yan",
                        "suffix": ""
                    }
                ],
                "year": 2005,
                "venue": "Speech and Audio Processing",
                "volume": "13",
                "issue": "",
                "pages": "330--344",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Gwo-Lang Yan. \"Speech act modeling and verification of spontaneous speech with disfluency in a spoken dialogue system\". In Speech and Audio Processing, IEEE Transactions on Volume 13, Issue 3, May 2005 Page(s):330 -344.",
                "links": null
            },
            "BIBREF20": {
                "ref_id": "b20",
                "title": "Edit Disfluency Detection and Correction Using a Cleanup Language Model and an Alignment Model",
                "authors": [
                    {
                        "first": "Jui-Feng",
                        "middle": [],
                        "last": "Yeh",
                        "suffix": ""
                    },
                    {
                        "first": "Chung-Hsien",
                        "middle": [],
                        "last": "Wu",
                        "suffix": ""
                    }
                ],
                "year": 2006,
                "venue": "Speech, and Language Processing",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Yeh, Jui-Feng and Chung-Hsien Wu.\"Edit Disfluency Detection and Correction Using a Cleanup Language Model and an Alignment Model,\"accepted by IEEE Trans. Audio, Speech, and Language Processing, 2006.",
                "links": null
            },
            "BIBREF21": {
                "ref_id": "b21",
                "title": "Association Pattern Language Modeling",
                "authors": [
                    {
                        "first": "Jen-Tzung",
                        "middle": [],
                        "last": "Chien",
                        "suffix": ""
                    }
                ],
                "year": 2005,
                "venue": "IEEE Transactions on Audio, Speech, and Language Processing",
                "volume": "99",
                "issue": "",
                "pages": "1--10",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Chien, Jen-Tzung,\"Association Pattern Language Modeling.\" IEEE Transactions on Audio, Speech, and Language Processing : Accepted for future publication Volume PP, Issue 99, 2005 Page(s):1-10",
                "links": null
            },
            "BIBREF22": {
                "ref_id": "b22",
                "title": "Conditional random fields: probabilistic models for segmenting and labeling sequence data",
                "authors": [
                    {
                        "first": "J",
                        "middle": [],
                        "last": "Lafferty",
                        "suffix": ""
                    },
                    {
                        "first": "A",
                        "middle": [],
                        "last": "Mccallum",
                        "suffix": ""
                    },
                    {
                        "first": "F",
                        "middle": [],
                        "last": "Pereira",
                        "suffix": ""
                    }
                ],
                "year": 2001,
                "venue": "ICML",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Lafferty, J., A. McCallum, and F. Pereira. \"Conditional random fields: probabilistic models for segmenting and labeling sequence data.\" In ICML, 2001.",
                "links": null
            },
            "BIBREF23": {
                "ref_id": "b23",
                "title": "The EARS Fall 2004 Rich Transcription Evaluation Plan",
                "authors": [],
                "year": 2004,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "The EARS Fall 2004 Rich Transcription Evaluation Plan August 30, 2004\u3002",
                "links": null
            }
        },
        "ref_entries": {
            "TABREF0": {
                "html": null,
                "num": null,
                "text": "\u505a\u70ba\u8a13\uf996\u4ee5\u53ca\u6e2c\u8a66\u8a9e\uf9be\u3002\u5176\u4e2d\u88ab\u4fee\u6b63\u8a5e(editing word)\u932f\u8aa4\uf961\u70ba 17.3%\uff0c\u76f8\u5c0d\u65bc DF-gram\u3001HMM\u3001\u6700\u5927\u71b5\u4ee5 \u53ca N-gram \u52a0\u6821\u6b63\u4e4b\u6df7\u5408\u6a21\u578b\u7684\u65b9\u6cd5\u5206\u5225\ufa09\u4f4e\uf9ba 11.7%\u30018.7%\u30018%\u4ee5\u53ca 3.9%\u3002\u5728\u7d66\u5b9a\u4e2d\u65b7\u9ede\u7684\u60c5\u6cc1\u4e0b\uff0c\u88ab ]\u3002John Bear \u61c9\u7528\uf967\u540c\u77e5\uf9fc\uf92d\u6e90\uf92d\u91dd\u5c0d\uf967\uf9ca \u66a2\u8a9e\uf9ca\u9032\ufa08\u5075\u6e2c\u53ca\u4fee\u6b63[8]\uff0cAnand Venkataraman \u4f7f\u7528\u4eba\u5de5\u8a02\u5b9a\u4e4b\u898f\u5247\uf92d\u5224\u65b7\uf967\uf9ca\u66a2\u8a9e\uf9ca[9]\u3001 Matthias Honal \uf9dd\u7528\u566a\u97f3\u983b\u9053(Noisy Channel)\u7684\u89c0\uf9a3\uff0c\u904b\u7528\uf967\u540c\u7279\u5fb5\u8a13\uf996\u51fa\u7d71\u8a08\u6a21\u578b\u4e26\u4ee5\u7dda\u6027\u7d44 \u5408\uf92d\u4fee\u6b63\u4e4b[10]\u3002Charniak and Johnson \u5efa\uf9f7\u3127\u57fa\u65bc\u8a5e\u6027\u7279\u5fb5\u4e4b\u5206\uf9d0\u5668\uf92d\u9810\u6e2c\u53ef\u88ab\u522a\u9664\u5340\u57df[11]\u3002 Nakatani and Hirschberg \uf9dd\u7528\u8072\u5b78\u3001\u97f3\u97fb\u5b78\u4ee5\u53ca\u8a9e\u8a00\u7279\u5fb5\u5efa\uf9f7\u3127\u6c7a\u7b56\u6a39\u6a21\u578b\uf92d\u5075\u6e2c\u91cd\u8907[12]\u3002",
                "type_str": "table",
                "content": "<table><tr><td colspan=\"7\">\u4fee\u6b63\u6a21\u578b (Cleanup Model): \u6839\u64da\u8a5e\u7d61\u3001\u53ef\u80fd\u7684\u4e2d\u65b7\u9ede\u4ee5\u53ca\u64f7\u53d6\u4e4b\u8a9e\u8a00\u7279\u5fb5\u914d\u5408\u5c0d\u61c9\u7684\uf96b\uf969\u5c0d\u8a5e \u7d61\u4f5c\u4fee\u6b63\u3002 \u6700\u5f8c\uff0c\u8a5e\u7d61\u7d93\u904e\u4fee\u6b63\u6a21\u578b\u4fee\u6b63\u4e4b\u5f8c\uff0c\u6211\u5011\u5c07\u5f97\u5230\u4e2d\u65b7\u9ede\u8cc7\u8a0a\u3001\u4fee\u6b63\u5f8c\u7d50\u679c\u3001\u4ee5\u53ca\u8fa8\uf9fc\u5f8c\u7d50\u679c\u3002\u800c \u7cfb\u7d71\uf9ca\u7a0b\u5206\u70ba\u8a13\uf996\u548c\u6e2c\u8a66\uf978\u90e8\u4efd\uff0c\u5206\u5225\u5982\u4e0b: \u8a13\uf996\u90e8\u4efd--\u9996\u5148\uff0c\u5f9e MCDC \u8a9e\uf9be\u4e2d\u7d93\u7531\u5b50\u7279\u5fb5\u5806\u5c0e\u5f97\u5230\u5b50\u7279\u5fb5\u3002\u9019\u4e9b\u5b50\u7279\u5fb5\u5247\u6210\u70ba\u6211\u5011\u4fee\u6b63\u6a21 \u7528\u6700\u5927\u71b5\u6a21\u578b\u4ee5\u53ca\u96b1\u85cf\u99ac\u53ef\u592b\u6a21\u578b\u4fee\u6b63\uf967\uf9ca\u66a2\u8a9e\uf9ca[7Snover \u7b49\u4eba\u4ee5\u53ca Joungbum \u7b49\u4eba\uf9dd\u7528\u8f49\u63db\u5b78\u7fd2(Transformation-Based Learning)\u5075\u6e2c\uf967\uf9ca\u66a2\u8a9e\uf9ca \u578b\u4e2d\u7684\u7279\u5fb5\u51fd\uf969\u3002\u6700\u5f8c\u5247\u662f\u5c0d\u8a9e\uf9be\u9032\ufa08\u8a9e\u8a00\u7279\u5fb5\u64f7\u53d6\u4e26\u5c0d\u6240\u6709\u7684\u7279\u5fb5\u51fd\uf969\u9032\ufa08\uf96b\uf969\u4f30\u6e2c\uff0c\u9019\u4e9b\uf96b</td></tr><tr><td colspan=\"7\">[13][14]\u3002\u65e5\u672c\u7684 Furui \u5247\u81f4\uf98a\u65bc\u53e3\u8a9e\u5316\u8a9e\u97f3\u8fa8\uf9fc\u4e4b\u7814\u7a76[15]\u3002\u570b\u5167\u65b9\u9762\uff0c\u4e2d\u7814\u9662\u91dd\u5c0d\uf967\uf9ca\u66a2\u8a9e\uf9ca \uf969\u5373\u70ba\u6e2c\u8a66\u6642\u4fee\u6b63\u6a21\u578b\u4e4b\uf96b\uf969\u3002</td></tr><tr><td colspan=\"7\">\u7684\u8a9e\u97f3\u7279\u6027\u505a\u5206\u6790[16]\uff0c\u53f0\u7063\u5927\u5b78\u7814\u7a76\u95dc\u65bc\uf967\uf9ca\u66a2\u8a9e\uf9ca\u4e2d\u65b7\u9ede\u5075\u6e2c\u4e4b\u7279\u5fb5[17]\u3001\u4ea4\u901a\u5927\u5b78\u96fb\u4fe1\u5de5 \u6e2c\u8a66\u90e8\u5206--\u8a9e\u97f3\u8a0a\u865f\u7d93\u7531\u8fa8\uf9fc\u5668\u9032\ufa08\u8a9e\u97f3\u8fa8\uf9fc\u5f8c\uff0c\u5f97\u5230\u97f3\u7bc0\u7d61\uff0c\u5c07\u6b64\u97f3\u7bc0\u7d61\u914d\u5408\u8a5e\u5178\u7d93\u7531\u69cb\u8fad\u5f8c</td></tr><tr><td colspan=\"7\">\u7a0b\u7cfb\u5247\u91dd\u5c0d\u81ea\u767c\u6027\u4e2d\u6587\u8a9e\u97f3\u5efa\uf9f7\u8fa8\uf9fc\u7cfb\u7d71[18]\u4ee5\u53ca\u81ea\u767c\u6027\u5c0d\u8a71\u8a9e\u97f3\u8fa8\uf9fc\u505a\u7814\u7a76[19]\u3002\u8fd1\uf98e\uf92d\uff0c\u6210 \u6703\u5f97\u5230\u8a5e\u7d61\uff0c\u7136\u5f8c\u5c0d\u6b64\u8a5e\u7d61\u505a\u8a9e\u8a00\u7279\u5fb5\u64f7\u53d6\u3002\u6700\u5f8c\uff0c\u4fee\u6b63\u6a21\u578b\u5247\u6839\u64da\u8a5e\u7d61\u3001\u53ef\u80fd\u7684\u4e2d\u65b7\u9ede\u4ee5\u53ca\u6240</td></tr><tr><td colspan=\"7\">\u529f\u5927\u5b78\u4e5f\u6295\u5165\u5927\uf97e\u7814\u7a76\u80fd\uf97e\u65bc\u53e3\u8a9e\u5c0d\u8a71\u7cfb\u7d71\u4e2d\uf967\uf9ca\u66a2\u8a9e\u97f3\u4e4b\u8a9e\u97f3\u52d5\u4f5c\u578b\u614b\u6a21\u578b\u5316\u8207\u9a57\u8b49[20]\u4ee5 \u64f7\u53d6\u4e4b\u8a9e\u8a00\u7279\u5fb5\uff0c\u914d\u5408\u8a13\uf996\u6240\u5f97\u5230\u7684\uf96b\uf969\u6a21\u578b\uff0c\u627e\u51fa\u4e2d\u65b7\u9ede\u8cc7\u8a0a\u3001\u4fee\u6b63\u5f8c\u7d50\u679c\u3001\u4ee5\u53ca\u8fa8\uf9fc\u5f8c\u7d50\u679c</td></tr><tr><td colspan=\"7\">\u53ca\u904b\u7528\u8a9e\u8a00\u6a21\u578b\u8207\u6821\u6b63\u6a21\u578b\uf92d\u5c0d\u7de8\u8f2f\uf967\uf9ca\u66a2\u8a9e\uf9ca\u505a\u4fee\u6b63[21]\u3002\u672c\u6587\u5247\u91dd\u5c0d\u7de8\u8f2f\uf967\uf9ca\u66a2\u8a9e\uf9ca\u63d0\u51fa\u4e00 \u4e26\u5c07\u5176\u8f38\u51fa\u3002 \u6458\u8981 \u91dd\u5c0d\u53e3\u8a9e\u5316\u8a9e\u97f3\u4e2d\u4e4b\uf967\uf9ca\u66a2\u8a9e\uf9ca(disfluency)\u73fe\u8c61\uff0c\u672c\u6587\u63d0\u51fa\u4ee5\uf967\u5b9a\u9577\ufa01\u7279\u5fb5\u4e4b\u689d\u4ef6\u96a8\u6a5f\u57df\u3002\uf9dd\u7528\uf9fa\u614b\u8f49\u79fb \uf9dd\u7528\u689d\u4ef6\u96a8\u6a5f\u57df\u4e4b\u4fee\u6b63\u65b9\u6cd5\u3002 \u800c\u4fee\u6b63\uf967\uf9ca\u66a2\u8a9e\uf9ca\u4e4b\uf9ca\u7a0b\u4ee5\u5f0f\u5b50(1)\u8868\u793a\uff0c\u4e00\u8a9e\u97f3\u8a0a\u865f X \u8f38\u5165\u5f8c\uff0c\u6211\u5011\u8981\u5f97\u5230\u5176\u76f8\u5c0d\u61c9\u4e4b\u6700\u4f73\uf9fa</td></tr><tr><td colspan=\"7\">\u7279\u5fb5\u51fd\uf969\u3001\u89c0\u6e2c\u7279\u5fb5\u51fd\uf969\u4ee5\u53ca\u76f8\u5c0d\u61c9\u4e4b\uf96b\uf969\uff0c\u91dd\u5c0d\uf967\uf9ca\u66a2\u8a9e\uf9ca\u9032\ufa08\u4fee\u6b63\u3002\u5176\u4e2d\u89c0\u6e2c\u7279\u5fb5\u51fd\uf969\u53ef\u6574\u5408\u591a\u7a2e\u77e5 \u614b\u5e8f\uf99c S\uff0c\u65bc\u662f\u5f15\u5165\u8a5e\u5e8f\uf99c W \u6b64\uf96b\uf969\uff0c\u4e4b\u5f8c\u5728\u689d\u4ef6\u7368\uf9f7\u7684\u5047\u8a2d\u4e0b\uff0c\u5f97\u5230\u6700\u5f8c\u7684\u5f0f\u5b50\uff0c\u4e5f\u5c31\u662f\u5f9e 2. \u7cfb\u7d71\u67b6\u69cb \u8a9e\u97f3\u8a0a\u865f\u6211\u5011\u53ef\u7d93\u904e\u8fa8\uf9fc\u5668\u5f97\u5230\u8a5e\u5e8f\uf99c\uff0c\u800c\u5f8c\u6211\u5011\u5f9e\u8a5e\u5e8f\uf99c\u627e\u5230\u76f8\u5c0d\u61c9\u7684\uf9fa\u614b\u5e8f\uf99c\u3002\u5728\u672c\uf941\u6587 \uf9fc\uf92d\u6e90\uff0c\u5305\u62ec\u524d\u5f8c\u6587\u76f8\u95dc\u7279\u5fb5\u3001\uf967\uf9ca\u66a2\u76f8\u95dc\u7279\u5fb5\u4ee5\u53ca\u5716\u6a23\u7b26\u5408\u76f8\u95dc\u7279\u5fb5\u3002\u5728\uf9fa\u614b\u65b9\u9762\u6211\u5011\u4f7f\u7528\u53ef\u8b8a\u52d5\u9577\ufa01 \u4e2d\uff0c\u6211\u5011\u4f7f\u7528\uf967\u5b9a\u9577\ufa01\u7279\u5fb5\u4e4b\u689d\u4ef6\u96a8\u6a5f\u57df\u5f97\u5230 ( ) S|W P \u3002 \u55ae\u4f4d\uff0c\u5305\u62ec\u8a5e\u3001\u5b57\u5143\uf905\u96c6(chunk)\u4ee5\u53ca\uf906\u5b50\u4e09\u7a2e\uf967\u540c\uf9fa\u614b\u3002\u5728\u8a55\u4f30\u4e0a\uff0c\u5247\u4f7f\u7528\u73fe\u4ee3\u6f22\u8a9e\u53e3\u8a9e\u5c0d\u8a71\u8a9e\uf9be\u5eab(MCDC) ( ) S argmax S| X P</td></tr><tr><td>\u015c</td><td/><td/><td/><td/><td/></tr><tr><td>arg max</td><td/><td>P</td><td>(</td><td colspan=\"3\">) ( S|W,X P W|X</td><td>)</td></tr><tr><td colspan=\"7\">\u4fee\u6b63\u8a5e\u932f\u8aa4\uf961\u70ba 6.1%\u3002\u5be6\u9a57\u8b49\u660e\u6240\u63d0\u4e4b\u6a21\u578b\u512a\u65bc\u5176\u4ed6\u65b9\u6cd5\uff0c\u4e26\u53ef\u6709\u6548\u5075\u6e2c\u4e26\u4fee\u6b63\u53e3\u8ff0\u8a9e\u8a00\u4e2d\u4e4b\uf967\uf9ca\u66a2\u8a9e\uf9ca\u3002 S W</td></tr><tr><td>S arg max</td><td>W</td><td>P</td><td colspan=\"3\">( ) ( S|W P W|X</td><td>)</td><td>1. \u7dd2\uf941</td></tr><tr><td colspan=\"7\">\u8981\u61c9\u7528\u8a9e\u97f3\u6280\u8853\u65bc\u4eba\u6a5f\u4ecb\u9762\u4e0a\uff0c\u8a9e\u97f3\u8fa8\uf9fc\u5247\u70ba\u6700\u91cd\u8981\u4e14\u6838\u5fc3\u4e4b\u6280\u8853\u4e4b\u4e00\u3002\u8fd1\u5341\uf98e\uf92d\uff0c\u8a9e\u97f3\u8fa8\u8a8d\u6280</td></tr><tr><td colspan=\"7\">\u8853\u5df2\u81fb\u65bc\u6210\u719f\u4e14\u84ec\u52c3\u767c\u5c55\u3002\u76ee\u524d\u7684\u8a9e\u97f3\u8fa8\u8a8d\u7cfb\u7d71\u5c0d\u65bc\uf929\uf95a\u7684\u8a9e\u97f3\u8f38\u5165\u8fa8\u8a8d\u6548\u679c\u6975\u4f73\uff0c\u7136\u800c\u8981\u5be6\u969b</td></tr><tr><td colspan=\"7\">\u61c9\u7528\uff0c\u5fc5\u9808\u8003\u616e\u53e3\u8a9e\u5316\u8a9e\u97f3[1]\u3002\u800c\u53e3\u8a9e\u5316\u8a9e\u97f3\u5e38\u6703\u4f34\u96a8\u8457\u975e\u6b63\u898f\u5316(ill-formed)\u4ee5\u53ca\uf967\uf9ca\u66a2\u8a9e\uf9ca</td></tr><tr><td colspan=\"7\">(disfluency)\uff0c\u9019\u4e9b\u73fe\u8c61\u6703\u9020\u6210\u76ee\u524d\u8fa8\uf9fc\u7cfb\u7d71\u7684\u932f\u8aa4\uf961\u5927\u5e45\ufa01\u63d0\u9ad8\uff0c\u4ee5\u81f3\u65bc\u7121\u6cd5\u61c9\u7528\u65bc\u65e5\u5e38\u751f\u6d3b</td></tr><tr><td colspan=\"7\">[2]\u3002\u800c\uf96b\u96dc\u8457\uf967\uf9ca\u66a2\u8a9e\uf9ca\u4e4b\u8fa8\uf9fc\u5f8c\u6587\u5b57\uff0c\u4e5f\u6703\u4f7f\u5f97\u4f7f\u7528\u8005\u6975\uf967\u5bb9\uf9e0\u95b1\uf95a\uff0c\u5c0d\u4f7f\u7528\u8005\u9020\u6210\u56f0\u64fe[3]\u3002</td></tr><tr><td colspan=\"7\">\u7de8\u8f2f\uf967\uf9ca\u66a2\u8a9e\uf9ca\u7d50\u69cb\u5171\u53ef\u5340\u5206\u70ba\u4ee5\u4e0b\u56db\u500b\u90e8\u4efd\u5982\u5716\u4e00\u6240\u793a\u3002</td></tr><tr><td/><td/><td colspan=\"3\">L</td><td colspan=\"2\">L L</td><td>LL</td><td>LL</td></tr><tr><td/><td/><td/><td/><td/><td colspan=\"2\">\u5716\u4e00 \u7de8\u8f2f\uf967\uf9ca\u66a2\u8a9e\uf9ca\u4e4b\u7d50\u69cb</td></tr><tr><td colspan=\"7\">\u7de8\u8f2f\uf967\uf9ca\u66a2\u8a9e\uf9ca\u5305\u62ec\u4e09\u7a2e\u578b\u5225\uff1a\u91cd\u8907(Repetition)\u3001\u4fee\u6b63(Repair) \u548c\u91cd\u958b\u59cb(Restart)\uff0c\u5176\u5b9a\u7fa9\u5982\u4e0b\u3002</td></tr><tr><td colspan=\"7\">\u91cd\u8907\u5373\u8a9e\u8005\u91cd\u8907\u8a9e\uf906\u7684\u67d0\u500b\u90e8\u4efd\uff0c\u4e5f\u5c31\u662f\u53ef\u522a\u9664\u5340\u57df\u8207\u4fee\u6b63\u5340\u57df\u7684\u8a9e\uf906\u91cd\u8907\u3002\u4fee\u6b63\u5373\u8a9e\u8005\u5c07\u8a9e\uf906</td></tr><tr><td colspan=\"7\">\u7684\u67d0\u500b\u90e8\u4efd\u505a\u4fee\u6b63\u3002\u4e5f\u5c31\u662f\u53ef\u522a\u9664\u5340\u57df\u5c07\u53d6\u4ee3\u4fee\u6b63\u5340\u57df\u4e26\u6539\u8b8a\u5b83\u7684\u610f\u601d\u3002\u91cd\u958b\u59cb:\u8a9e\u8005\u5c07\u672a\u5b8c\u6210</td></tr><tr><td colspan=\"7\">\u7684\u8a9e\uf906\u4e2d\u65b7\u4e26\u91cd\u65b0\u958b\u59cb\u53e6\u4e00\uf906\u3002\u4e5f\u5c31\u662f\u4e2d\u65b7\u9ede\u524d\u9762\u7684\u90e8\u5206\u5168\ufa26\u662f\u53ef\u522a\u9664\u5340\u57df\u3002</td></tr><tr><td colspan=\"7\">\u76f8\u95dc\u7684\u7814\u7a76\u5728\u570b\u5916\u65b9\u9762\uff0c ISCI \u4ee5\u53ca SRI \u7b49\u570b\u969b\u7814\u7a76\u4e2d\u5fc3\uf9dd\u7528\u8a9e\u8a00\u6a21\u578b\u4ee5\u53ca\u97fb\uf9d8\u6a21\u578b\u5075\u6e2c\uf967\uf9ca</td></tr><tr><td colspan=\"7\">\u66a2\u8a9e\uf9ca[4]\u3001\u7d50\u5408\u57fa\u65bc\u8a5e\u548c\u8a5e\u6027\u7684\u8a9e\u8a00\u6a21\u578b\u89e3\u6c7a\u91cd\u8907[5]\u548c\u4f7f\u7528\u96b1\u85cf\u4e8b\u4ef6\u8a9e\u8a00\u6a21\u578b\u76f4\u63a5\u5c0d\uf967\uf9ca\u66a2\u8a9e</td></tr><tr><td colspan=\"7\">\uf9ca\u9032\ufa08\u7d71\u8a08\u5f0f\u5206\u6790\u4ee5\u53ca\uf9dd\u7528\uf967\uf9ca\u66a2\u8a9e\uf9ca\u8a9e\u8a00\u6a21\u578b(DF-gram)\uf92d\u9810\u6e2c\u662f\u5426\u51fa\u73fe\uf967\uf9ca\u66a2\u73fe\u8c61[6]\u4ee5\u53ca\u4f7f</td></tr></table>"
            },
            "TABREF3": {
                "html": null,
                "num": null,
                "text": "\u4f3c\u7136\ufa01\u51fd\uf969\u7d93\u504f\u5fae\u5206\u5f8c\u8a2d\u70ba\uf9b2\uf92d\u89e3\u51fa\uf96b\uf969 \u0398 \uff0c\u672a\u5fc5\u70ba\u5c01\u9589\u89e3\u3002\u6545\u672c\u6587\u63a1\u7528\u4e00\u4e9b\u53cd\u8986(iterative)\u5f62 \u5f0f\u7684\u6280\u5de7\u53d6\u5f97\u4f7f log-\u4f3c\u7136\ufa01\u6700\u5927\u4e4b\uf96b\uf969\u3002\u56e0 IIS \u5177\u6709\u8f03\u5feb\u6536\u6b5b\u4e4b\u512a\u9ede\uff0c\u6545\u672c\uf941\u6587\u662f\uf9dd\u7528 IIS \u6f14\u7b97 \u6cd5\uf92d\u9032\ufa08\uf96b\uf969\u4f30\u6e2c\u3002IIS \u6f14\u7b97\u6cd5\u662f\u4ee5 GIS \u6f14\u7b97\u6cd5\u70ba\u57fa\u790e\u6539\u8b8a\u800c\u6210\uff0c\u5176\u512a\u9ede\u70ba\u6536\u6582\u901f\ufa01\u8f03 GIS \u6f14\u7b97 \u6cd5\u5feb\u3002\u6211\u5011\u4ee5\u6b64\u70ba\u57fa\u790e\u914d\u5408 Lafferty \u7b49\u4eba\u63d0\u51fa\u7684\u52d5\u614b\u898f\u5283\u6cd5\uf92d\u4f30\u7b97\uf96b\uf969\u3002",
                "type_str": "table",
                "content": "<table><tr><td colspan=\"17\">\u5728(24)\u5f0f\u4ee5\u53ca(25)\u5f0f\u4e2d\uff0cS \u70ba\u67d0\u4e00\u8a13\uf996\u8cc7\uf9be\u5176\u5305\u542b\u4e4b\u7279\u5fb5\u51fd\uf969\u7684\u7e3d\uf969\u5728\u5168\u90e8\u8a13\uf996\u8cc7\uf9be\u4e2d\u6700\u5927\u7684\u3002</td></tr><tr><td>S</td><td colspan=\"2\">=</td><td colspan=\"3\">s max</td><td>\u239b \u239c \u239c \u239d</td><td colspan=\"4\">t \u2211\u2211\u2211 k p,q</td><td>( f s k</td><td>( ) ( ) t 1 t p q ,s , \u2212</td><td colspan=\"2\">W</td><td>)</td><td>+</td><td>t \u2211\u2211\u2211 k p</td><td>( ) t p g s , ( k</td><td>W</td><td>)</td><td>\u239e \u239f \u239f \u23a0</td><td>(26)</td></tr><tr><td colspan=\"17\">Lafferty \u7b49\u4eba[24]\u89c0\u5bdf\u5230\u5c0d\u65bc\u4e00\u500b\u93c8\uf9fa\u7d50\u69cb\u7684\u689d\u4ef6\u96a8\u6a5f\u57df(CRFs)\uff0c\u7d66\u4e88\u89c0\u6e2c\u5e8f\uf99c W \u6240\u5f97\u5230\uf9fa \u614b\u5e8f\uf99c s \u7684\u689d\u4ef6\u6a5f\uf961 p(s|W)\u53ef\u7c21\u55ae\u7684\u7528\u77e9\u9663\u7684\u5f62\u5f0f\uf92d\u8868\u793a\u3002\u5c0d\u65bc\u89c0\u6e2c\u5e8f\uf99c W \u4e2d\u7684\u6bcf\u4e00\u500b\u4f4d\u7f6e t\uff0c ( ) [ ] , k p W s E f % \u70ba\u7279\u5fb5\u51fd\uf969 k f \u5176\u8a13\uf996\u8cc7\uf9be\u5206\u4f48\u7684\u671f\u671b\u503c:</td></tr><tr><td colspan=\"17\">\u6211\u5011\u5206\u5225\u5b9a\u7fa9\uf9ba\u4e00\u500b \u03ba \u03ba \u00d7 \u7684\u77e9\u9663\u96a8\u6a5f\u8b8a\uf969 ( ) ( ) ( P W n 1 l t 1 t M k kp q t p W ,s W ,s t 1 p,q 1 E f ,s f s ,s ,W ( ) W ) + \u2212 = = = \u23a1 \u23a4 \u23a3 \u23a6 \u2211 \u2211 \u2211 % %</td><td>= \u23a1 \u23a3</td><td>( M s s W ' , | t</td><td>)</td><td>\u23a4 \u23a6 \uff0c\u03ba\u70ba\uf9fa\u614b\u7684\u7a2e\uf9d0\u500b\uf969\u3002 (27)</td></tr><tr><td colspan=\"14\">( | , s',s |W ) [ ] ) = k p sW M t ( E f \u0398 % \u70ba\u9810\u4f30\u6e2c\u5206\u4f48\u7684\u671f\u671b\u503c: ( t 1 k k p k p,q exp f s \u03bb \u2212 \u239b \u239c = \u239c \u239d \u2211\u2211</td><td colspan=\"3\">s',s</td><td>t q</td><td>=</td><td>s,W</td><td>)</td><td>+</td><td>k \u2211\u2211 p</td><td>\u03bc</td><td>( g s k k</td><td>t p</td><td>=</td><td>s,W</td><td>)</td><td>\u239e \u239f \u239f \u23a0</td><td>(18)</td></tr><tr><td colspan=\"17\">\u6bcf\u500b \u6a5f\uf961 ( p s|W , M t W \u53ef\u8996\u70ba\u8868\u793a\u5728\u6642\u9593 t \u6642\uff0c\u6a21\u578b\u4e2d\u6bcf\u500b\u8f49\u79fb\u7684\u6b0a\u91cd\u3002\u65bc\u662f\u6211\u5011\u53ef\u4ee5\u5c07\u672a\u6b63\u898f\u5316\u7684\u689d\u4ef6 ( ) ( ) * P s|W \u8868\u793a\u70ba\u77e9\u9663\u7684\uf99a\u4e58\u7a4d: ) ( ) ( ) ( ) P W n 1 l t 1 t k k p q W ,s t 1 p,q 1 E f Ps | W f s , s, W + \u2212 \u0398 = = = \u23a1 \u23a4 \u23a3 \u23a6 \u2211 \u2211\u2211 % % (28)</td></tr><tr><td colspan=\"17\">( P s|W * Lafferty \u7b49\u4eba\u63d0\u51fa\u7684\u52d5\u614b\u898f\u5283\u6cd5(dynamic programming)\u5373\u662f\uf9dd\u7528 ( ) ( ) n 1 t 1 t t p q t 1 M s ,s |W + \u2212 P s W \u53ef\u8868\u793a\u70ba\u77e9\u9663 (19) ) | \u7684\u5f62\u5f0f\uff0c\u6545\u5f0f\u5b50(28)\u53ef\u8868\u793a\u70ba: = \u220f = \u800c\u6b63\u898f\u5316\u4fc2\uf969 Z(W)\uff0c\u53ea\u548c\u89c0\u6e2c\u5e8f\uf99c W \u6709\u95dc\uff0c\u70ba\u9577\ufa01 n+1 \u6642\u6240\u6709\u53ef\u80fd\u4e4b\uf9fa\u614b\u5e8f\uf99c\u7d44\u5408: ( ) ( ) ( ) ( n 1 l ) P W t 1 t k k p q p s|W , E f Ps | W f s , s, W</td><td>( ) M W t</td></tr><tr><td/><td/><td/><td/><td/><td/><td/><td/><td/><td colspan=\"2\">W ,s</td><td/><td/><td colspan=\"4\">t 1 p,q 1</td></tr><tr><td colspan=\"4\">( ) Z W</td><td>=</td><td>(</td><td colspan=\"11\">( ) ( ) 2 M W M W 1 W t1p , q1s ' , s ( ) n1 M W + L ( ) n 1 l P W f s ) start,stop ( t 1 k p</td><td>n 1 t 1 + = \u220f t q s',s \u23a1 = \u23a2 \u23a2 \u23a3</td><td>( ) M W t ) s,W</td><td>\u23a4 \u23a5 \u23a5 \u23a6</td><td>start,stop</td><td>(20)</td></tr><tr><td colspan=\"17\">\u6545\u6b63\u898f\u5316\u5f8c\u7684\u689d\u4ef6\u6a5f\uf961 p(s|W)\u53ef\u8868\u793a\u70ba: ( ) ( ) ( ) Z W n 1 t 1 t t p q t 1 M s ,s |W P s |W \u2212 = = \u220f + ( ) ( ) ( ) t t s' |W M s',s |W s |W t1 ( ) Z W</td><td>(21)</td></tr><tr><td colspan=\"17\">\u6211\u5011\uf9dd\u7528\u53cd\u8986(iterative)\u7684\u5f62\u5f0f\uff0c\u6bcf\u4e00\u56de\u5408\uf901\u65b0\u4e00\u6b21\uf96b\uf969:</td></tr><tr><td colspan=\"8\">k \u03bb \u03bb k = +\u0394</td><td colspan=\"3\">k \u03bb</td><td/><td/><td/><td/><td/><td>(22)</td></tr><tr><td colspan=\"2\">\u03bc</td><td>k</td><td>=</td><td colspan=\"2\">\u03bc</td><td>k</td><td colspan=\"2\">+\u0394</td><td>\u03bc</td><td>k</td><td/><td/><td/><td/><td/><td>(23)</td></tr><tr><td colspan=\"12\">\u5176\u4e2d\u6bcf\u4e00\u56de\u5408\uf901\u65b0\u7684\u503c</td><td/><td/><td/><td/></tr><tr><td colspan=\"17\">W k f s ,s ,W t 1 t p q f s ,s , t 1 t k p q n 1 l t 1 p,q 1 k f \u2212 + \u2212 = = \u0398 t 1 p,q 1 n l p s|W , p W p s |W , k p W ,s W ,s W ,s p ,s E f E = = \u2202 \u0398 = k L \u03bb \u2202 \u2212 \u0398 = \u2212 \u23a1 \u23a4 \u23a1 \u23a4 \u23a3 \u23a6 \u23a3 \u23a6 \u2211 \u2211\u2211 \u2211 % % ( ) ( ) k p s|W , k k p W ,s E f 1 log S E f \u03bb \u0398 \u23a1 \u23a4 \u23a3 \u23a6 \u0394 = \u23a1 \u23a4 % \u23a3 \u23a6 % \u2211\u2211 % ( ) ( ) k p s|W , k k p W ,s E g 1 log S E g \u03bc \u0398 \u23a1 \u23a4 \u23a3 \u23a6 % \u0394 = \u23a1 \u23a4 % \u23a3 \u23a6</td><td>(17) (24)</td><td>(25)</td></tr><tr><td colspan=\"17\">\uf974\u8981\u6c42\u5f97\u5168\u57df\u6700\u5927\u503c\u5247\u9808\uf9a8\u5f0f\u5b50(17)\u70ba\uf9b2\uff0c\u6c42\u5f97 \u0398 \u89e3\u3002\uf967\u904e\u4e00\u822c\uf92d\uf96f\uff0c\u9019\u662f\uf967\u53ef\ufa08\u7684\uff0c\u56e0\u70ba\u5c07 log-</td></tr></table>"
            }
        }
    }
}