| { | |
| "paper_id": "O14-5004", | |
| "header": { | |
| "generated_with": "S2ORC 1.0.0", | |
| "date_generated": "2023-01-19T08:04:34.482089Z" | |
| }, | |
| "title": "\u4f7f\u7528\u6982\u5ff5\u8cc7\u8a0a\u65bc\u4e2d\u6587\u5927\u8a5e\u5f59\u9023\u7e8c\u8a9e\u97f3\u8fa8\u8b58\u4e4b\u7814\u7a76 \u4f7f\u7528\u6982\u5ff5\u8cc7\u8a0a\u65bc\u4e2d\u6587\u5927\u8a5e\u5f59\u9023\u7e8c\u8a9e\u97f3\u8fa8\u8b58\u4e4b\u7814\u7a76 \u4f7f\u7528\u6982\u5ff5\u8cc7\u8a0a\u65bc\u4e2d\u6587\u5927\u8a5e\u5f59\u9023\u7e8c\u8a9e\u97f3\u8fa8\u8b58\u4e4b\u7814\u7a76 \u4f7f\u7528\u6982\u5ff5\u8cc7\u8a0a\u65bc\u4e2d\u6587\u5927\u8a5e\u5f59\u9023\u7e8c\u8a9e\u97f3\u8fa8\u8b58\u4e4b\u7814\u7a76 Exploring Concept Information for Mandarin Large Vocabulary Continuous Speech Recognition \u90dd\u67cf\u7ff0 \u90dd\u67cf\u7ff0 \u90dd\u67cf\u7ff0 \u90dd\u67cf\u7ff0\u3001 \u3001 \u3001 \u3001\u9673\u601d\u6f84 \u9673\u601d\u6f84 \u9673\u601d\u6f84 \u9673\u601d\u6f84\u3001 \u3001 \u3001 \u3001\u9673\u67cf\u7433 \u9673\u67cf\u7433 \u9673\u67cf\u7433 \u9673\u67cf\u7433", | |
| "authors": [ | |
| { | |
| "first": "Po-Han", | |
| "middle": [], | |
| "last": "Hao", | |
| "suffix": "", | |
| "affiliation": { | |
| "laboratory": "", | |
| "institution": "National Taiwan Normal University", | |
| "location": {} | |
| }, | |
| "email": "" | |
| }, | |
| { | |
| "first": "Ssu-Cheng", | |
| "middle": [], | |
| "last": "Chen", | |
| "suffix": "", | |
| "affiliation": { | |
| "laboratory": "", | |
| "institution": "National Taiwan Normal University", | |
| "location": {} | |
| }, | |
| "email": "" | |
| }, | |
| { | |
| "first": "Berlin", | |
| "middle": [], | |
| "last": "Chen", | |
| "suffix": "", | |
| "affiliation": { | |
| "laboratory": "", | |
| "institution": "National Taiwan Normal University", | |
| "location": {} | |
| }, | |
| "email": "berlin@csie.ntnu.edu.tw" | |
| } | |
| ], | |
| "year": "", | |
| "venue": null, | |
| "identifiers": {}, | |
| "abstract": "Language modeling (LM) is part and parcel of automatic speech recognition (ASR), since it can assist ASR to constrain the acoustic analysis, guide the search through multiple candidate word strings, and quantify the acceptability of the final output hypothesis given an input utterance. This paper investigates and develops language model adaptation techniques for use in ASR and its main contribution is twofold. First, we propose a novel concept language modeling (CLM) approach to rendering the relationships between a search history and an upcoming word. Second, the instantiations of CLM are constructed with different levels of lexical granularities, such as words and document clusters. In addition, we also explore the incorporation of word proximity cues into the model formulation of CLM, getting around the \"bag-of-words\" assumption. A series of experiments conducted on a Mandarin large vocabulary continuous speech recognition (LVCSR) task demonstrate that our proposed language models can offer substantial improvements over the baseline N-gram system, and achieve performance competitive to, or better than, some state-of-the-art language model adaptation methods.", | |
| "pdf_parse": { | |
| "paper_id": "O14-5004", | |
| "_pdf_hash": "", | |
| "abstract": [ | |
| { | |
| "text": "Language modeling (LM) is part and parcel of automatic speech recognition (ASR), since it can assist ASR to constrain the acoustic analysis, guide the search through multiple candidate word strings, and quantify the acceptability of the final output hypothesis given an input utterance. This paper investigates and develops language model adaptation techniques for use in ASR and its main contribution is twofold. First, we propose a novel concept language modeling (CLM) approach to rendering the relationships between a search history and an upcoming word. Second, the instantiations of CLM are constructed with different levels of lexical granularities, such as words and document clusters. In addition, we also explore the incorporation of word proximity cues into the model formulation of CLM, getting around the \"bag-of-words\" assumption. A series of experiments conducted on a Mandarin large vocabulary continuous speech recognition (LVCSR) task demonstrate that our proposed language models can offer substantial improvements over the baseline N-gram system, and achieve performance competitive to, or better than, some state-of-the-art language model adaptation methods.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Abstract", | |
| "sec_num": null | |
| } | |
| ], | |
| "body_text": [ | |
| { | |
| "text": "\u6587\u4ef6\u6458\u8981\u7b49\u5404\u7a2e\u4efb\u52d9\u4e4b\u4e2d\uff0c\u4e26\u6210\u70ba\u95dc\u9375\u7684\u7d44\u6210 (Rosenfeld, 2000; Bellegarda, 2004) (Furui et al., 2012; O'Shaughnessy et al., 2013) (Kuhn, 1988 )\uff0c\u4ee5\u53ca\u6e90\u81ea\u65bc\u8cc7\u8a0a\u6aa2\u7d22\u9818\u57df\u7684\u4e3b\u984c\u6a21\u578b(Topic \u4f7f\u7528\u6982\u5ff5\u8cc7\u8a0a\u65bc\u4e2d\u6587\u5927\u8a5e\u5f59\u9023\u7e8c\u8a9e\u97f3\u8fa8\u8b58\u4e4b\u7814\u7a76 49 Model) (Blei & Lafferty, 2009) \u7b49\uff1b\u800c\u4e3b\u984c\u6a21\u578b\u5728\u8a9e\u97f3\u8fa8\u8b58\u4efb\u52d9\u7684\u5be6\u4f5c\u4e0a\uff0c\u53c8\u4ee5\u6a5f\u7387\u5f0f\u6f5b\u85cf \u8a9e\u610f\u5206\u6790(Probabilistic Latent Semantic Analysis, PLSA) (Hofmann, 1999) \u4ee5\u53ca\u5176\u5ef6\u4f38\u72c4\u5229 \u514b\u91cc\u5206\u914d(Latent Dirichlet Allocation, LDA) (Blei et al., 2003) (Kuhn, 1988) \uff0c\u7528\u5728\u8a9e\u97f3\u8fa8\u8b58\u904e\u7a0b\u4e2d (Lau et al., 1993; Troncoso & Kawahara, 2005) (Hofmann, 1999 )\u4ee5\u53ca\u5176\u5ef6\u4f38\u72c4\u5229\u514b\u91cc\u5206\u914d (Latent Dirichlet Allocation, LDA) (Blei et al., 2003) (Blei, 2014; Kim et al., 2013; Potapenko & Konstantin, 2013 (Ortmanns et al., 1997) (Mikolov et al., 2010; Deng & Yu, 2014) \uff0c\u4f86\u5be6\u73fe\u6982\u5ff5\u8a9e\u8a00\u6a21\u578b \u6240\u6b32\u64f7\u53d6\u7684\u8a5e\u5f59\u548c\u8a9e\u610f\u4f7f\u7528\u8cc7\u8a0a\u3002\u540c\u6642\uff0c\u6211\u5011\u4ea6\u5e0c\u671b\u80fd\u5c07\u5176\u5b83\u5728\u8cc7\u8a0a\u6aa2\u7d22\u9818\u57df\u4ee5\u767c\u5c55\u76f8 \u7576\u4e0d\u932f\u7684\u65b0\u7a4e\u8a9e\u8a00\u6a21\u578b (Blei, 2014; Chen et al., 2004; Kim et al., 2013; Zhai, 2008 ", | |
| "cite_spans": [ | |
| { | |
| "start": 21, | |
| "end": 38, | |
| "text": "(Rosenfeld, 2000;", | |
| "ref_id": "BIBREF21" | |
| }, | |
| { | |
| "start": 39, | |
| "end": 56, | |
| "text": "Bellegarda, 2004)", | |
| "ref_id": "BIBREF1" | |
| }, | |
| { | |
| "start": 57, | |
| "end": 77, | |
| "text": "(Furui et al., 2012;", | |
| "ref_id": "BIBREF10" | |
| }, | |
| { | |
| "start": 78, | |
| "end": 105, | |
| "text": "O'Shaughnessy et al., 2013)", | |
| "ref_id": "BIBREF19" | |
| }, | |
| { | |
| "start": 106, | |
| "end": 117, | |
| "text": "(Kuhn, 1988", | |
| "ref_id": "BIBREF13" | |
| }, | |
| { | |
| "start": 175, | |
| "end": 198, | |
| "text": "(Blei & Lafferty, 2009)", | |
| "ref_id": "BIBREF3" | |
| }, | |
| { | |
| "start": 277, | |
| "end": 292, | |
| "text": "(Hofmann, 1999)", | |
| "ref_id": "BIBREF11" | |
| }, | |
| { | |
| "start": 340, | |
| "end": 359, | |
| "text": "(Blei et al., 2003)", | |
| "ref_id": "BIBREF4" | |
| }, | |
| { | |
| "start": 360, | |
| "end": 372, | |
| "text": "(Kuhn, 1988)", | |
| "ref_id": "BIBREF13" | |
| }, | |
| { | |
| "start": 384, | |
| "end": 402, | |
| "text": "(Lau et al., 1993;", | |
| "ref_id": "BIBREF15" | |
| }, | |
| { | |
| "start": 403, | |
| "end": 429, | |
| "text": "Troncoso & Kawahara, 2005)", | |
| "ref_id": "BIBREF24" | |
| }, | |
| { | |
| "start": 430, | |
| "end": 444, | |
| "text": "(Hofmann, 1999", | |
| "ref_id": "BIBREF11" | |
| }, | |
| { | |
| "start": 493, | |
| "end": 512, | |
| "text": "(Blei et al., 2003)", | |
| "ref_id": "BIBREF4" | |
| }, | |
| { | |
| "start": 513, | |
| "end": 525, | |
| "text": "(Blei, 2014;", | |
| "ref_id": "BIBREF2" | |
| }, | |
| { | |
| "start": 526, | |
| "end": 543, | |
| "text": "Kim et al., 2013;", | |
| "ref_id": "BIBREF12" | |
| }, | |
| { | |
| "start": 544, | |
| "end": 572, | |
| "text": "Potapenko & Konstantin, 2013", | |
| "ref_id": "BIBREF20" | |
| }, | |
| { | |
| "start": 573, | |
| "end": 596, | |
| "text": "(Ortmanns et al., 1997)", | |
| "ref_id": "BIBREF18" | |
| }, | |
| { | |
| "start": 597, | |
| "end": 619, | |
| "text": "(Mikolov et al., 2010;", | |
| "ref_id": "BIBREF17" | |
| }, | |
| { | |
| "start": 620, | |
| "end": 636, | |
| "text": "Deng & Yu, 2014)", | |
| "ref_id": "BIBREF8" | |
| }, | |
| { | |
| "start": 698, | |
| "end": 710, | |
| "text": "(Blei, 2014;", | |
| "ref_id": "BIBREF2" | |
| }, | |
| { | |
| "start": 711, | |
| "end": 729, | |
| "text": "Chen et al., 2004;", | |
| "ref_id": "BIBREF5" | |
| }, | |
| { | |
| "start": 730, | |
| "end": 747, | |
| "text": "Kim et al., 2013;", | |
| "ref_id": "BIBREF12" | |
| }, | |
| { | |
| "start": 748, | |
| "end": 758, | |
| "text": "Zhai, 2008", | |
| "ref_id": "BIBREF26" | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "\u8a9e\u8a00\u6a21\u578b(Language Models, LM)\u5df2\u88ab\u5ee3\u6cdb\u5730\u4f7f\u7528\u65bc\u8a9e\u97f3\u8fa8\u8b58\u3001\u6a5f\u5668\u7ffb\u8b6f\u3001\u8cc7\u8a0a\u6aa2\u7d22\u4ee5\u53ca", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "\u52d5\u614b\u4f86\u8f14\u52a9\u6216\u8abf\u6574 N \u9023\u8a9e\u8a00\u6a21\u578b\u65bc\u9810\u6e2c\u8a5e\u5f59\u51fa\u73fe\u7684\u6a5f\u7387\u3002\u5176\u57fa\u672c\u6982\u5ff5\u662f\u5982\u679c\u6211\u5011\u8b1b\u4e86\u4e00 \u4e9b\u8a5e\u5f59\uff0c\u5247\u4e00\u6bb5\u6642\u9593\u5167\u9019\u4e9b\u8a5e\u5f59\u518d\u6b21\u51fa\u73fe\u7684\u6a5f\u7387\u6703\u5f88\u9ad8\u3002\u6211\u5011\u56e0\u6b64\u53ef\u4ee5\u5229\u7528\u6b64\u7dda\u7d22\u5728\u8a9e \u97f3\u8fa8\u8b58\u904e\u7a0b\u4e2d\u4e0d\u65b7\u5730\u7522\u751f\u4e00\u500b\u8a9e\u8a00\u6a21\u578b(\u4f8b\u5982\u55ae\u9023\u5feb\u53d6\u6a21\u578b)\uff0c\u4e26\u900f\u904e\u7dda\u6027\u7d44\u5408\u7684\u65b9\u5f0f\u8207 \u539f\u59cb N \u9023\u8a9e\u8a00\u6a21\u578b(\u4f8b\u5982\u4e09\u9023\u8a9e\u8a00\u6a21\u578b)\u7d50\u5408\u4f86\u52d5\u614b\u5730\u8abf\u9069\u8a9e\u97f3\u8fa8\u8b58\u6240\u9700\u7684\u8a9e\u8a00\u6a21\u578b\uff1a ( ) i i i i i i i i i H H w n w w w P w w w P , ) 1 ( ) | ( ) | ( 1 2 Trigram 1 2 Trigram \u22c5 \u2212 + \u22c5 = \u2212 \u2212 \u2212 \u2212 \u03bb \u03bb (1) \u5176\u4e2d i H \u4ee3\u8868\u8a5e\u5f59 i w \u5c0d\u61c9\u7684\u6b77\u53f2\u8a5e\u5e8f\u5217 i H \u4e2d\u7684\u7e3d\u8a5e\u6578\uff1b ( ) i i H w n , \u662f i w \u5728 i H \u51fa\u73fe\u7684\u6b21 \u6578\u3002\u904e\u53bb\u8a31\u591a\u7814\u7a76\u4ea6\u5be6\u9a57\u4e86\u4e8c\u9023\u5feb\u53d6(", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "\u8a9e\u8a00\u6a21\u578b(Language Models, LM)\u5df2\u88ab\u5ee3\u6cdb\u5730\u4f7f\u7528\u65bc\u8a9e\u97f3\u8fa8\u8b58\u3001\u6a5f\u5668\u7ffb\u8b6f\u3001\u8cc7\u8a0a\u6aa2\u7d22\u4ee5\u53ca", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": ") | ( x y w w P \u3002\u89f8\u767c \u5c0d\u6a21\u578b\u904b\u7528\u65bc\u8a9e\u8a00\u6a21\u578b\u6642\uff0c\u662f\u7531\u5f85\u9810\u6e2c\u8a5e\u5f59 i w \u5c0d\u61c9\u7684\u6b77\u53f2\u8a5e\u5e8f\u5217 i H \u4e2d\u5c0b\u627e\u8a5e\u5f59 i w \u7684\u53ef \u80fd\u7684\u89f8\u767c\u9805 i L h h h , , , 2 1 \u22ef (\u5047\u8a2d\u6b77\u53f2\u8a5e\u5e8f\u5217 i L i h h h H , , , 2 1 \u22ef = \uff0c\u800c\u6bcf\u4e00\u500b\u6b77\u53f2\u8a5e\u5f59 l h \u5c0d \u65bc\u8a5e\u5f59 i w \u7684\u89f8\u767c\u6a5f\u7387\u70ba ) | ( l i h w P )\uff0c\u4e26\u5c07\u9019\u4e9b\u89f8\u767c\u9805\u5206\u5225\u9810\u6e2c\u7684\u689d\u4ef6\u6a5f\u7387 ) | ( l i h w P \u52d5\u614b \u7dda\u6027\u7d44\u5408\u800c\u6210\u70ba\u89f8\u767c\u5c0d\u6a21\u578b\uff1a \u2211 \u2212 = \u2212 = 1 1 Trigger ) | ( 1 1 ) | ( i L l l i i i i h w P L H w P (2) \u800c\u5f0f(2)\u52d5\u614b\u7522\u751f\u7684\u89f8\u767c\u5c0d\u6a21\u578b\u4ea6\u53ef\u518d\u900f\u904e\u7dda\u6027\u7d44\u5408\u65b9\u5f0f\u8207\u539f\u59cb N \u9023\u8a9e\u8a00\u6a21\u578b\u7d50\u5408\u4f86\u52d5 \u614b\u8abf\u9069\u8a9e\u97f3\u8fa8\u8b58\u6240\u9700\u7684\u8a9e\u8a00\u6a21\u578b(\u5982\u5f0f(1)\u7684\u7d50\u5408\u65b9\u5f0f)\u3002 2.3 \u4e3b\u984c\u6a21\u578b \u4e3b\u984c\u6a21\u578b \u4e3b\u984c\u6a21\u578b \u4e3b\u984c\u6a21\u578b \u901a\u5e38\u5728\u8cc7\u8a0a\u6aa2\u7d22\u4efb\u52d9\u4e0a\uff0c\u4e3b\u984c\u6a21\u578b\u85c9\u7531\u4e00\u7d44\u6f5b\u85cf\u4e3b\u984c\u5206\u5e03\u7528\u4f86\u63cf\u8ff0\"\u8a5e\u5f59-\u6587\u4ef6\"\u5171\u540c\u51fa \u73fe\u7684\u7279\u6027(Blei & Lafferty, 2009)\u3002\u7576\u4e3b\u984c\u6a21\u578b\u88ab\u61c9\u7528\u81f3\u8a9e\u97f3\u8fa8\u8b58\u904e\u7a0b\u6642\uff0c\u5f85\u9810\u6e2c\u8a5e\u5f59 i w \u8207 \u5176\u5c0d\u61c9\u6b77\u53f2\u8a5e\u5e8f\u5217 i H (\u5728\u6b64\u53ef\u8996\u70ba\u4e00\u7bc7\u6587\u4ef6)\u4e4b\u76f8\u4e92\u95dc\u4fc2\u5176\u6709\u4e00\u7d44\u6f5b\u85cf\u7684\u4e3b\u984c\u5206\u5e03\u7528\u4f86 \u63cf\u8ff0\u6b77\u53f2\u8a5e\u5e8f\u5217 i H \u8207\u5f85\u9810\u6e2c\u8a5e\u5f59 i w \u5171\u540c\u51fa\u73fe\u95dc\u4fc2\uff0c\u4e0d\u518d\u662f\u55ae\u7d14\u5730\u7d93\u7531\u8a08\u7b97 i w \u5728 i H \u7684 \u51fa\u73fe\u983b\u7387\u800c\u4f30\u6e2c\uff0c\u800c\u662f\u900f\u904e i w \u51fa\u73fe\u5728\u4e0d\u540c\u6f5b\u85cf\u4e3b\u984c\u5206\u5e03\u7684\u983b\u7387\u4ee5\u53ca i H \u7522\u751f\u9019\u4e9b\u6f5b\u85cf\u4e3b \u984c\u7684\u53ef\u80fd\u6027\u4f86\u6c7a\u5b9a\uff0c\u662f\u67d0\u7a2e\u7a0b\u5ea6\u4e0a\u7684\u6982\u5ff5\u6bd4\u5c0d(Concept Matching)\u3002\u6a5f\u7387\u5f0f\u6f5b\u85cf\u8a9e\u610f\u5206\u6790 (Probabilistic Latent Semantic Analysis, PLSA)", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "\u8a9e\u8a00\u6a21\u578b(Language Models, LM)\u5df2\u88ab\u5ee3\u6cdb\u5730\u4f7f\u7528\u65bc\u8a9e\u97f3\u8fa8\u8b58\u3001\u6a5f\u5668\u7ffb\u8b6f\u3001\u8cc7\u8a0a\u6aa2\u7d22\u4ee5\u53ca", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "\u662f\u6700\u5e38\u88ab\u4f7f\u7528\u7684\u4e3b\u984c\u6a21\u578b\u5be6\u4f8b\u3002\u5728\u6b64\u8209 \u6a5f\u7387\u5f0f\u6f5b\u85cf\u8a9e\u610f\u5206\u6790\u70ba\u4f8b\u4f86\u4f5c\u8aaa\u660e\uff0c\u7576\u5176\u88ab\u7528\u81f3\u8a9e\u97f3\u8fa8\u8b58\u4f86\u9032\u884c\u8a9e\u8a00\u6a21\u578b\u8abf\u9069\u6642\uff0c\u57fa\u65bc \u6b77\u53f2\u8a5e\u5e8f\u5217 i H \u4f86\u9810\u6e2c\u8a5e\u5f59 i w \u7684\u767c\u751f\u6a5f\u7387\u53ef\u8868\u793a\u70ba(Gildea & Hofmann, 1999)\uff1a ) | ( ) | ( ) | ( 1 PLSA i k K k k i i i H T P T w P H w P \u2211 = = (3) \u5176\u4e2d k T \u70ba\u67d0\u4e00\u500b\u6f5b\u5728\u4e3b\u984c\uff0c\u800c ) | ( k i T w P \u8207 ) | ( i k H T P \u5206\u5225\u8868\u793a\u8a5e\u5f59 i w \u767c\u751f\u5728\u4e3b\u984c k T \u7684 \u6a5f\u7387\u4ee5\u53ca\u6b77\u53f2\u8a5e\u5e8f\u5217 i H \u7522\u751f\u6b64\u4e3b\u984c\u7684\u6a5f\u7387\u3002\u6211\u5011\u5047\u8a2d\u6bcf\u4e00\u500b\u6f5b\u85cf\u4e3b\u984c\u7522\u751f\u5019\u9078\u8a5e\u7684\u6a5f \u7387 ) | ( k i T w P \u4e0d\u56e0\u8a5e\u5e8f\u5217\u641c\u5c0b\u53ca\u62d3\u5c55\u904e\u7a0b\u800c\u8b8a\u52d5\uff0c\u53ef\u5148\u85c9\u7531\u6700\u5927\u5316\u8abf\u9069(\u6216\u8a13\u7df4)\u8a9e\u6599\u767c\u751f \u6a5f\u7387\u800c\u6c42\u5f97\uff1b\u4f46\u7531\u65bc\u6b77\u53f2\u8a5e\u5e8f\u5217\u5728\u8a9e\u97f3\u8fa8\u8b58\u4e4b\u524d\u4e0d\u80fd\u4e8b\u5148\u6c7a\u5b9a\uff0c\u800c\u4e14\u6578\u91cf\u975e\u5e38\u591a\u4e26\u4e14\u6703 \u96a8\u8a9e\u97f3\u8fa8\u8b58\u904e\u7a0b\u6f14\u9032\u800c\u6539\u8b8a\uff0c\u6bcf\u4e00\u500b\u6b77\u53f2\u8a5e\u5e8f\u5217\u5c0d\u65bc\u4e3b\u984c\u5206\u5e03\u7684\u6b0a\u91cd\u5fc5\u9808\u5728\u8a9e\u97f3\u8fa8\u8b58\u904e \u7a0b\u4f7f\u7528\u671f\u671b\u503c\u6700\u5927\u5316(Expectation Maximization, EM)\u6f14\u7b97\u6cd5(Dempster, 1977)\u4f86\u9032\u884c\u7dda\u4e0a \u4f7f\u7528\u6982\u5ff5\u8cc7\u8a0a\u65bc\u4e2d\u6587\u5927\u8a5e\u5f59\u9023\u7e8c\u8a9e\u97f3\u8fa8\u8b58\u4e4b\u7814\u7a76 51 (\u52d5\u614b)\u4f30\u6e2c\u3002\u6a5f\u7387\u5f0f\u6f5b\u85cf\u8a9e\u610f\u5206\u6790\u7684\u512a\u9ede\u662f\u5728\u6c7a\u5b9a\u5f85\u9810\u6e2c\u8a5e\u5f59 i w \u767c\u751f\u7684\u6a5f\u7387\u6642\uff0c\u4e0d\u50c5\u6703 \u8003\u616e\u6574\u500b\u6b77\u53f2\u8a5e\u5e8f\u5217 i H \u7684\u4e3b\u984c\u5206\u5e03\u7279\u6027\uff0c\u800c\u4e14\u6703\u96a8\u8a9e\u97f3\u8fa8\u8b58\u5019\u9078\u8a5e\u5e8f\u5217\u641c\u5c0b\u7684\u6f14\u9032\uff0c \u52d5\u614b\u8abf\u6574\u8a5e\u5e8f\u5217\u6240\u542b\u6709\u7684\u6f5b\u85cf\u4e3b\u984c\u5206\u5e03\u8cc7\u8a0a\u3002\u800c\u5f0f(3)\u52d5\u614b\u5730\u7522\u751f\u7684\u6a5f\u7387\u5f0f\u6f5b\u85cf\u8a9e\u610f\u5206\u6790 \u6a21\u578b\u4ea6\u53ef\u518d\u900f\u904e\u7dda\u6027\u7d44\u5408\u65b9\u5f0f\u8207\u539f\u59cb N \u9023\u8a9e\u8a00\u6a21\u578b\u7d50\u5408\u4f86\u52d5\u614b\u8abf\u9069\u8a9e\u97f3\u8fa8\u8b58\u6240\u9700\u7684\u8a9e\u8a00 \u6a21\u578b(\u5982\u5f0f(1)\u7684\u7d50\u5408\u65b9\u5f0f)\u3002 \u53e6\u4e00\u65b9\u9762\uff0c\u72c4\u5229\u514b\u91cc\u5206\u914d\u64c1\u6709\u8207\u6a5f\u7387\u5f0f\u6f5b\u85cf\u8a9e\u610f\u5206\u6790\u76f8\u4f3c\u7684\u6578\u5b78\u8868\u793a\u5f0f\uff0c\u53ef\u8996\u70ba\u5f8c \u8005\u4e4b\u5ef6\u4f38\uff0c\u800c\u4e14\u72c4\u5229\u514b\u91cc\u5206\u914d\u5728\u8a31\u591a\u8a9e\u97f3\u8fa8\u8b58\u4efb\u52d9\u4e0a\u90fd\u5c55\u73fe\u4e86\u4e0d\u932f\u7684\u6548\u7528(Tam & Schultz, 2005)\u3002\u5169\u500b\u6a21\u578b\u9593\u7684\u4e3b\u8981\u5dee\u7570\u5728\u65bc\u6a5f\u7387\u5f0f\u6f5b\u85cf\u8a9e\u610f\u5206\u6790\u5047\u8a2d\u5176\u6a21\u578b\u53c3\u6578\u5728\u53c3\u6578\u7a7a\u9593\u4e0a\u662f \u56fa\u5b9a\u548c\u672a\u77e5\u5411\u91cf\uff0c\u800c\u72c4\u5229\u514b\u91cc\u5206\u914d\u5c0d\u65bc\u6a21\u578b\u53c3\u6578\u591a\u4e86\u5148\u5099\u9650\u5236(a Priori Constraints)\uff0c\u8a8d\u70ba \u53c3\u6578\u5411\u91cf\u672c\u8eab\u4e5f\u662f\u96a8\u6a5f\u8b8a\u6578\uff0c\u9075\u5faa\u8457\u67d0\u7a2e\u72c4\u5229\u514b\u91cc\u5206\u5e03\u7279\u6027\u3002\u7531\u65bc\u72c4\u5229\u514b\u91cc\u5206\u914d\u6a21\u578b\u7684 \u6700\u4f73\u5316\u8f03\u70ba\u56f0\u96e3\u3001\u4e0d\u5bb9\u6613\u9054\u5230\u6b63\u78ba\u7684\u4f30\u6e2c\uff0c\u8a31\u591a\u8fd1\u4f3c\u7684\u4f30\u6e2c\u6f14\u7b97\u6cd5\u50cf\u662f\u8b8a\u52d5\u6027\u8c9d\u6c0f\u8fd1\u4f3c (Variational Approximation)\u6f14\u7b97\u6cd5\u6216\u662f\u5409\u535c\u68ee\u53d6\u6a23(Gibbs Sampling)\u6f14\u7b97\u6cd5\u56e0\u6b64\u88ab\u63d0\u51fa\u4f86 \u4f30\u6e2c\u72c4\u5229\u514b\u91cc\u5206\u914d\u4e4b\u6a21\u578b\u53c3\u6578(Blei & Lafferty, 2009)\u3002\u95dc\u65bc\u4e3b\u984c\u6a21\u578b\u7684\u56de\u9867\u8207\u8fd1\u671f\u767c\u5c55\uff0c \u53ef\u4ee5\u53c3\u8003", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "\u8a9e\u8a00\u6a21\u578b(Language Models, LM)\u5df2\u88ab\u5ee3\u6cdb\u5730\u4f7f\u7528\u65bc\u8a9e\u97f3\u8fa8\u8b58\u3001\u6a5f\u5668\u7ffb\u8b6f\u3001\u8cc7\u8a0a\u6aa2\u7d22\u4ee5\u53ca", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "\u2211 \u2032 \u220f \u2032 \u2211 \u220f = \u2211 \u2032 \u2032 \u2211 = = \u2208 \u2032 = \u2032 \u2032 \u2208 = \u2208 \u2032 \u2208 c c c c c L l l c L l l i c i c i i i i i i i W c", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "\u8a9e\u8a00\u6a21\u578b(Language Models, LM)\u5df2\u88ab\u5ee3\u6cdb\u5730\u4f7f\u7528\u65bc\u8a9e\u97f3\u8fa8\u8b58\u3001\u6a5f\u5668\u7ffb\u8b6f\u3001\u8cc7\u8a0a\u6aa2\u7d22\u4ee5\u53ca", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "i i ) | ( ) | ( ) | ( ) | ( ) | ( ) | ( ) | ( ) | ( ) | , ( ) | ( ) | , ( ) , | ( 1 1 WCLM (4) \u5176 W \u4ee3\u8868\u8a9e\u8005\u6240\u8b1b\u8a9e\u53e5\u6240\u6b32\u8868\u9054\u7684\u8a9e\u8a00\u8cc7\u8a0a\uff0c\u5728\u6b64\u6211\u5011\u5148\u4ee5\u8a9e\u97f3\u8fa8\u8b58\u521d\u6b65(\u7b2c\u4e00\u968e\u6bb5)\u6240 \u7522\u751f\u7684\u8a5e\u5716(Word Graph)", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "\u8a9e\u8a00\u6a21\u578b(Language Models, LM)\u5df2\u88ab\u5ee3\u6cdb\u5730\u4f7f\u7528\u65bc\u8a9e\u97f3\u8fa8\u8b58\u3001\u6a5f\u5668\u7ffb\u8b6f\u3001\u8cc7\u8a0a\u6aa2\u7d22\u4ee5\u53ca", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "EQUATION", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [ | |
| { | |
| "start": 0, | |
| "end": 8, | |
| "text": "EQUATION", | |
| "ref_id": "EQREF", | |
| "raw_str": "\u4f86\u8fd1\u4f3c(\u8a5e\u5716\u5305\u542b\u6240\u6709\u53ef\u80fd\u7684\u5019\u9078\u8a5e\u5e8f\u5217)\uff1b \u800c c \u4ee3\u8868\u8207 W \u6240\u6b32\u8868\u9054\u7684\u8a9e\u8a00\u8cc7\u8a0a\u6709\u95dc\u7684\u4e00\u7d44\u6982\u5ff5\u95dc\u9375\u8a5e\u7d44\u3002\u5f9e\u5f0f(4)\u7684\u63a8\u5c0e\u53ef\u770b\u51fa\u8a5e\u6982 \u5ff5\u8a9e\u8a00\u6a21\u578b\u6b32\u6a21\u578b\u5316(\u7d00\u9304)\u7576\u67d0\u500b\u6982\u5ff5\u95dc\u9375\u8a5e c \u51fa\u73fe\u7684\u60c5\u6cc1\u4e0b\uff0c\u5f85\u9810\u6e2c\u8a5e\u5f59 i w \u8207\u5176\u6b77\u53f2 \u8a5e\u5e8f\u5217 i H \u5171\u540c\u51fa\u73fe\u7684\u95dc\u4fc2\u3002\u540c\u6642\uff0c\u8003\u91cf\u6a21\u578b\u4f30\u6e2c\u4e4b\u53ef\u884c\u6027\uff0c\u5f0f(4)\u9032\u4e00\u6b65\u5047\u8a2d\u7576\u67d0\u4e00\u500b \u6982\u5ff5\u95dc\u9375\u8a5e c \u51fa\u73fe\u7684\u60c5\u6cc1\u4e0b\uff0c\u5f85\u9810\u6e2c\u8a5e\u5f59 i w \u8207\u5176\u6b77\u53f2\u8a5e\u5e8f\u5217 i H \u4e2d\u4efb\u610f\u7684\u8a5e\u5f59\u4e4b\u9593\u662f\u5f7c \u6b64\u7368\u7acb\u7684\uff0c\u4e5f\u5c31\u662f\u6240\u8b02\u7684\u8a5e\u888b(Bag-of-Words)\u5047\u8a2d\u3002\u800c\u5f0f(4)\u4e2d ) | ( c w P i \u8207 ) | ( c h P l \u53ef\u5f9e \u8abf\u9069\u8a9e\u6599\u5eab\u88e1\u6982\u5ff5\u95dc\u9375\u8a5e c \u6240\u51fa\u73fe\u8655\u7684\u9130\u8fd1\u8cc7\u8a0a(Proximity Information)\uff0c\u6216\u8005\u8aaa\u662f\u51fa\u73fe \u8655\u4e0a\u4e0b\u6587\u7684\u8a5e\u5f59\u5206\u5e03\u800c\u4f30\u6e2c\u5f97\uff1b ) | ( W c P \u53ef\u900f\u904e\u9069\u7576\u65b9\u5f0f\u8a08\u7b97 W \u8207 c \u4e4b\u76f8\u4f3c\u5ea6\u800c\u6c42\u5f97\u3002 \u5be6\u52d9\u4e0a\uff0c\u6211\u5011\u9996\u5148\u906d\u9047\u5230\u7684\u554f\u984c\u5c31\u662f\u300c\u5982\u4f55\u6311\u9078\u5177\u4ee3\u8868\u6027\u7684\u95dc\u9375\u8a5e\u7d44\uff1f\u300d\u3002\u70ba\u6b64\uff0c \u672c\u8ad6\u6587\u5728\u6311\u9078\u6982\u5ff5\u95dc\u9375\u8a5e\u6642\u904b\u7528\u4e86\u5169\u968e\u6bb5\u7684\u6311\u9078\u65b9\u5f0f\uff0c\u5982\u5716 1 \u6240\u793a\u3002\u5728\u7b2c\u4e00\u968e\u6bb5\u6642\uff0c\u6211 \u5011 \u5229 \u7528 \u4e86 \u5728 \u8cc7 \u8a0a \u6aa2 \u7d22 \u9818 \u57df \u4e4b \u4e2d \u5e38 \u4f7f \u7528 \u7684 \u865b \u64ec \u95dc \u806f \u56de \u994b (Pseudo-Relevance Feedback, PRF)(Baeza-Yates & Ribeiro-Neto, 2011) \uff0c \u4e26 \u5229 \u7528 \u57fa \u65bc \u5eab \u723e \u8c9d \u514b \u2500 \u840a \u4f2f \u52d2 \u5dee \u7570 \u91cf (Kullback-Leibler Divergence, KL-Divergence) \u4e4b \u67e5 \u8a62 \u8207 \u6587 \u4ef6 \u6a21 \u578b \u5316 \u6280 \u8853 (Kullback & Leibler, 1951; Zhai, 2008)\uff0c\u4ee5\u8a5e\u5716 W(\u542b\u6709\u6b32\u8868\u9054\u7684\u8a5e\u5f59\u548c\u8a9e\u610f\u8cc7\u8a0a)\u70ba\u67e5\u8a62\u5f9e\u8abf\u9069\u8a9e\u6599\u7684 \u6587\u4ef6\u96c6\u6aa2\u7d22\u51fa\u4e00\u7d44\u8f03\u70ba\u76f8\u95dc\u7684\u6587\u4ef6\u5b50\u96c6\uff0c\u7a31\u9019\u4e9b\u6587\u4ef6\u70ba\u865b\u64ec\u95dc\u806f\u6587\u4ef6(Pseudo-Relevance Documents)\uff0c\u4e26\u5047\u8a2d\u9019\u4e9b\u6587\u4ef6\u542b\u6709\u8207\u6240\u6b32\u8868\u9054\u7684\u8a9e\u8a00\u8cc7\u8a0a\u6709\u95dc\u7684\u6982\u5ff5\u3002 \u5728\u7b2c\u4e8c\u968e\u6bb5\u6642\uff0c\u6211\u5011\u9032\u4e00\u6b65\u5f9e\u865b\u64ec\u95dc\u806f\u6587\u4ef6\u5b50\u96c6\u88e1\u6311\u9078\u51fa\u4e00\u7d44\u4e00\u5b9a\u6578\u91cf\u7684\u6982\u5ff5\u95dc\u9375 \u8a5e\u7d44\uff0c\u7136\u5f8c\u85c9\u7531\u9019\u7d44\u6982\u5ff5\u95dc\u9375\u8a5e\u7d44\u4f86\u91cf\u5316(\u6a5f\u7387\u5316)\u6b77\u53f2\u8a5e\u5e8f\u5217\u4e2d\u6240\u6709\u8a5e\u5f59\u8207\u5f85\u9810\u6e2c\u8a5e\u5f59 \u5728\u6b64\u6982\u5ff5\u95dc\u9375\u8a5e\u7d44\u4e0b\u7684\u5171\u540c\u51fa\u73fe\u95dc\u4fc2\u3002\u95dc\u65bc\u6982\u5ff5\u95dc\u9375\u8a5e\u6311\u9078\u6e96\u5247\uff0c\u6211\u5011\u53ef\u4ee5\u57fa\u65bc\u8a5e\u983b\u8207 \u53cd\u5411\u6587\u4ef6\u983b\u7387\u5206\u6578(TF-IDF Score)(Baeza-Yates & Ribeiro-Neto, 2011)\u3002\u8a5e\u983b\u8207\u53cd\u5411\u6587\u4ef6\u983b \u7387\u5206\u6578\u662f\u4e00\u9805\u5e38\u88ab\u7528\u65bc\u8cc7\u8a0a\u6aa2\u7d22\u4ee5\u53ca\u6587\u5b57\u5206\u6790\u9818\u57df\u4e2d\u7684\u6280\u8853\uff0c\u5176\u516c\u5f0f\u53ef\u4ee5\u8868\u793a\u5982\u4e0b\uff1a \uf8f3 \uf8f2 \uf8f1 > \u00d7 + = ohterwise f if n N f w m j j m j m j 0 0 ) / log( ) log 1 ( , , ,", | |
| "eq_num": "(5)" | |
| } | |
| ], | |
| "section": "\u8a9e\u8a00\u6a21\u578b(Language Models, LM)\u5df2\u88ab\u5ee3\u6cdb\u5730\u4f7f\u7528\u65bc\u8a9e\u97f3\u8fa8\u8b58\u3001\u6a5f\u5668\u7ffb\u8b6f\u3001\u8cc7\u8a0a\u6aa2\u7d22\u4ee5\u53ca", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "\u4f7f\u7528\u6982\u5ff5\u8cc7\u8a0a\u65bc\u4e2d\u6587\u5927\u8a5e\u5f59\u9023\u7e8c\u8a9e\u97f3\u8fa8\u8b58\u4e4b\u7814\u7a76 \u5716 \u5716 \u5716 \u5716 1. \u4e0a\u8ff0\u7684\u8a5e\u983b\u8207\u53cd\u5411\u6587\u4ef6\u983b\u7387\u5206\u6578\u4e3b\u8981\u53ef\u5206\u70ba\u5169\u500b\u4e3b\u8981\u90e8\u5206 \u5176\u4e2d\u7684 m j f , \u5247\u4ee3\u8868\u8a5e\u5f59 j w \u5728\u6b64\u6587\u4ef6 TF)\uff0c\u53ef\u4ee5\u89e3\u91cb\u70ba\u5177\u8d8a\u9ad8\u8a5e\u983b\u7684\u8a5e\u5f59\u5c0d\u6587\u4ef6\u4f86\u8b1b\u8d8a\u91cd\u8981 \u4e4b\u5247\u662f\u4ee3\u8868\u8a5e\u5f59 j w \u51fa\u73fe\u5728\u6240\u6709\u865b\u64ec\u95dc\u806f\u6587\u4ef6\u7684\u6587\u4ef6\u500b\u6578 Document Frequency, IDF)\uff0c\u7576\u67d0\u4e00\u8a5e\u5f59\u51fa\u73fe\u50c5\u51fa\u73fe\u5728\u5c11\u6578\u7684\u6587\u4ef6\u4e4b\u4e2d \u7368\u7279\u6027\u3002\u6211\u5011\u671f\u671b\u900f\u904e\u5f0f(5)\u80fd\u627e\u51fa\u5177\u6709\u91cd\u8981\u6027\u8207\u7368\u7279\u6027\u7684\u8a5e\u5f59\u505a\u70ba\u6982\u5ff5\u95dc\u9375\u8a5e 3.2 \u4ee5\u7fa4\u805a\u9762\u5411\u5efa\u7acb\u6982\u5ff5\u8a9e\u8a00\u6a21\u578b \u4ee5\u7fa4\u805a\u9762\u5411\u5efa\u7acb\u6982\u5ff5\u8a9e\u8a00\u6a21\u578b \u4ee5\u7fa4\u805a\u9762\u5411\u5efa\u7acb\u6982\u5ff5\u8a9e\u8a00\u6a21\u578b \u4ee5\u7fa4\u805a\u9762\u5411\u5efa\u7acb\u6982\u5ff5\u8a9e\u8a00\u6a21\u578b \u7fa4\u805a\u6982\u5ff5\u8a9e\u8a00\u6a21\u578b(Cluster-based Concept Language Model, CCLM) \u4ef6\u96c6\u5167\u4e4b\u6587\u4ef6\u53ef\u4ee5\u7531\u4e00\u7d44\u6982\u5ff5\u985e\u5225 \u8207\u9019\u4e9b\u6982\u5ff5\u985e\u5225\u7684\u500b\u5225\u95dc\u806f\u7a0b\u5ea6\u4f86\u7372\u5f97\u8a9e\u53e5\u53ef\u80fd\u7684\u6982\u5ff5\u5206\u5e03 \u64da\uff1a \u2211 \u2211 = \u2211 \u2211 = \u2208 \u2208 C C C C C i i W H w P 1 - CCLM ) , | ( \u5176 \u4e2d \u6982 \u5ff5 \u985e \u5225 \u7684 \u6c42 \u53d6 \u53ef \u900f \u904e \u4e00 \u822c \u5206 \u7fa4 \u6f14 \u7b97 \u6cd5 \u8af8 \u5982 Ribeiro-Neto, 2011)\u800c\u6c42\u5f97\uff1b (C P \u91cf\u5f62\u5f0f\uff0c\u8a08\u7b97 W \u8207 C \u4e4b(\u9918\u5f26)\u76f8\u4f3c\u5ea6\u800c\u6c42\u5f97 \u9023\u8a9e\u8a00\u6a21\u578b\u6a5f\u7387\uff0c\u53ef\u900f\u904e\u6700\u5927\u5316\u76f8\u4f3c\u6a5f\u7387\u4f30\u6e2c\u800c\u5f97 \u805a\u6982\u5ff5\u8a9e\u8a00\u6a21\u578b\u6b32\u6a21\u578b\u5316(\u7d00\u9304) \u6b77\u53f2\u8a5e\u5e8f\u5217 i H \u5171\u540c\u51fa\u73fe\u7684\u95dc\u4fc2 \u4f7f\u7528\u6982\u5ff5\u8cc7\u8a0a\u65bc\u4e2d\u6587\u5927\u8a5e\u5f59\u9023\u7e8c\u8a9e\u97f3\u8fa8\u8b58\u4e4b\u7814\u7a76 1. \u8a5e\u6982\u5ff5\u8a9e\u8a00\u8a9e\u8a00\u6a21\u578b\u6d41\u7a0b\u5716 \u8a5e\u6982\u5ff5\u8a9e\u8a00\u8a9e\u8a00\u6a21\u578b\u6d41\u7a0b\u5716 \u8a5e\u6982\u5ff5\u8a9e\u8a00\u8a9e\u8a00\u6a21\u578b\u6d41\u7a0b\u5716 \u8a5e\u6982\u5ff5\u8a9e\u8a00\u8a9e\u8a00\u6a21\u578b\u6d41\u7a0b\u5716 \u4e0a\u8ff0\u7684\u8a5e\u983b\u8207\u53cd\u5411\u6587\u4ef6\u983b\u7387\u5206\u6578\u4e3b\u8981\u53ef\u5206\u70ba\u5169\u500b\u4e3b\u8981\u90e8\u5206\uff1a\u7b2c\u4e00\u90e8\u5206\u70ba log 1 ( + \u5728\u6b64\u6587\u4ef6 m d \u4e2d\u6240\u51fa\u73fe\u7684\u6b21\u6578\uff0c\u7a31\u4e4b\u70ba\u8a5e\u983b(Term Frequency, \u53ef\u4ee5\u89e3\u91cb\u70ba\u5177\u8d8a\u9ad8\u8a5e\u983b\u7684\u8a5e\u5f59\u5c0d\u6587\u4ef6\u4f86\u8b1b\u8d8a\u91cd\u8981\uff1b\u7b2c\u4e8c\u90e8\u5206\u70ba ) / log( j n N \uff0c \u51fa\u73fe\u5728\u6240\u6709\u865b\u64ec\u95dc\u806f\u6587\u4ef6\u7684\u6587\u4ef6\u500b\u6578\uff0c\u7a31\u4e4b\u70ba\u53cd\u5411\u6587\u4ef6\u983b\u7387 \u7576\u67d0\u4e00\u8a5e\u5f59\u51fa\u73fe\u50c5\u51fa\u73fe\u5728\u5c11\u6578\u7684\u6587\u4ef6\u4e4b\u4e2d\uff0c\u5247\u6b64\u8a5e\u5f59\u8d8a\u5177\u6709 \u80fd\u627e\u51fa\u5177\u6709\u91cd\u8981\u6027\u8207\u7368\u7279\u6027\u7684\u8a5e\u5f59\u505a\u70ba\u6982\u5ff5\u95dc\u9375\u8a5e\u3002 \u4ee5\u7fa4\u805a\u9762\u5411\u5efa\u7acb\u6982\u5ff5\u8a9e\u8a00\u6a21\u578b \u4ee5\u7fa4\u805a\u9762\u5411\u5efa\u7acb\u6982\u5ff5\u8a9e\u8a00\u6a21\u578b \u4ee5\u7fa4\u805a\u9762\u5411\u5efa\u7acb\u6982\u5ff5\u8a9e\u8a00\u6a21\u578b \u4ee5\u7fa4\u805a\u9762\u5411\u5efa\u7acb\u6982\u5ff5\u8a9e\u8a00\u6a21\u578b based Concept Language Model, CCLM)\u5047\u8a2d\u5728\u8abf\u9069\u8a9e\u6599\u7684\u6587 \u4ef6\u96c6\u5167\u4e4b\u6587\u4ef6\u53ef\u4ee5\u7531\u4e00\u7d44\u6982\u5ff5\u985e\u5225 C \u4f86\u8868\u793a\uff0c\u85c9\u7531\u8a9e\u8005\u8b1b\u7684\u8a9e\u53e5\u6240\u6b32\u8868\u9054\u7684\u8a9e\u8a00\u8cc7\u8a0a \u8207\u9019\u4e9b\u6982\u5ff5\u985e\u5225\u7684\u500b\u5225\u95dc\u806f\u7a0b\u5ea6\u4f86\u7372\u5f97\u8a9e\u53e5\u53ef\u80fd\u7684\u6982\u5ff5\u5206\u5e03\uff0c\u4e26\u505a\u70ba\u8a9e\u8a00\u6a21\u578b\u9810\u6e2c\u7684\u6839 \u220f \u2032 \u2032 \u220f \u2032 \u2032 \u2208 \u2032 = \u2032 \u2032 = \u2208 \u2032 \u2208 C C C C L l l L l l i C i i i i i W C P C h P W C P C h P C w P W C P C H P W C P C H w P 1 1 ) | ( ) | ( ) | ( ) | ( ) | ( ) | ( ) | ( ) | ( ) | , ( \u5176 \u4e2d \u6982 \u5ff5 \u985e \u5225 \u7684 \u6c42 \u53d6 \u53ef \u900f \u904e \u4e00 \u822c \u5206 \u7fa4 \u6f14 \u7b97 \u6cd5 \u8af8 \u5982 K-Means \u6f14 \u7b97 \u6cd5 (Baeza-Yates & ) | W \u53ef\u57fa\u65bc\u5c07\u8a9e\u8a00\u8cc7\u8a0a W \u8207\u6bcf\u4e00\u500b\u6982\u5ff5\u985e\u5225 C \u8868\u793a\u6210\u5411 \u76f8\u4f3c\u5ea6\u800c\u6c42\u5f97\uff1b ) | ( C w P i \u4ee3\u8868\u6982\u5ff5\u985e\u5225 C \u9810\u6e2c\u8a5e\u5f59 \u53ef\u900f\u904e\u6700\u5927\u5316\u76f8\u4f3c\u6a5f\u7387\u4f30\u6e2c\u800c\u5f97(Zhai, 2008)\u3002\u5f9e\u5f0f(6)\u7684\u63a8\u5c0e\u53ef\u770b\u51fa\u7fa4 )\u7576\u67d0\u4e00\u500b\u6982\u5ff5\u985e\u5225 C \u51fa\u73fe\u7684\u60c5\u6cc1\u4e0b\uff0c\u5f85\u9810\u6e2c\u8a5e\u5f59 \u5171\u540c\u51fa\u73fe\u7684\u95dc\u4fc2\u3002 53 ) ,m j f \uff0c (Term Frequency, \uff0c\u5176\u4e2d j n \u7a31\u4e4b\u70ba\u53cd\u5411\u6587\u4ef6\u983b\u7387(Inverse \u5247\u6b64\u8a5e\u5f59\u8d8a\u5177\u6709 \u3002 \u5047\u8a2d\u5728\u8abf\u9069\u8a9e\u6599\u7684\u6587 \u85c9\u7531\u8a9e\u8005\u8b1b\u7684\u8a9e\u53e5\u6240\u6b32\u8868\u9054\u7684\u8a9e\u8a00\u8cc7\u8a0a W \u4e26\u505a\u70ba\u8a9e\u8a00\u6a21\u578b\u9810\u6e2c\u7684\u6839 (6) Yates & \u8868\u793a\u6210\u5411 \u9810\u6e2c\u8a5e\u5f59 i w \u7684\u55ae \u7684\u63a8\u5c0e\u53ef\u770b\u51fa\u7fa4 \u6e2c\u8a5e\u5f59 i w \u8207\u5176 54 \u90dd\u67cf\u7ff0 \u7b49 \u5716 \u5716 \u5716 \u5716 2. \u7fa4\u805a\u6982\u5ff5\u8a9e\u8a00\u6a21\u578b\u793a\u610f\u5716 \u7fa4\u805a\u6982\u5ff5\u8a9e\u8a00\u6a21\u578b\u793a\u610f\u5716 \u7fa4\u805a\u6982\u5ff5\u8a9e\u8a00\u6a21\u578b\u793a\u610f\u5716 \u7fa4\u805a\u6982\u5ff5\u8a9e\u8a00\u6a21\u578b\u793a\u610f\u5716 \u6211\u5011\u53ef\u4ee5\u5c07\u5f0f(6)\u4e2d\u6982\u5ff5\u985e\u5225 C \u9810\u6e2c\u8a5e\u5f59 i w \u7684\u8a9e\u8a00\u6a21\u578b\u5ef6\u4f38\u6210\u70ba\u96d9\u9023(Bigram)\u6216\u8005 \u4e09\u9023(Trigram)\u8a9e\u8a00\u6a21\u578b\uff0c\u800c\u53ef\u5206\u5225\u5f97\u5230\u4e0b\u9762\u5169\u500b\u8868\u793a\u5f0f\uff1a \u2211 \u220f \u2032 \u2032 \u2032 \u2211 \u220f = \u2208 \u2032 = \u2032 \u2212 \u2032 \u2032 \u2208 = \u2212 C C C L l l l C L l l l L i i i i i W C P C h h P C h P W C P C h h P C h P C h w P W H w P 2 1 1 2 1 1 2 - CCLM ) | ( ) , | ( ) | ( ) | ( ) , | ( ) | ( ) , | ( ) , | ( (7) \u2211 \u220f \u2032 \u2032 \u2032 \u2032 \u2211 \u220f = \u2208 \u2032 = \u2032 \u2212 \u2212 \u2208 = \u2212 \u2212 \u2212 C C C L l l l l C L l l l l L L i i i i i W C P C h h h P C h h P C h P W C P C h h h P C h h P C h P C h h w P W H w P 3 1 2 1 2 1 3 1 2 1 2 1 1 3 - CCLM ) | ( ) , , | ( ) , | ( ) | ( ) | ( ) , , | ( ) , | ( ) | ( ) , , | ( ) , | ( (8) \u5982 \u6b64 \u4e00 \u4f86 \uff0c \u6982 \u5ff5 \u8a9e \u8a00 \u6a21 \u578b \u53ef \u4ee5 \u540c \u6642 \u8003 \u616e \u8a5e \u5f59 \u9593 \u51fa \u73fe \u7684 \u5148 \u5f8c \u898f \u5247 \u6027 \u6216 \u8005 \u662f \u9130 \u8fd1 \u8cc7 \u8a0a (Proximity Information)\uff0c\u53ef\u4ee5\u514d\u9664\u4ee5\u8a5e\u888b(Bag-of-Words)\u5047\u8a2d\u7684\u9650\u5236\u3002\u6700\u5f8c\uff0c\u5f0f(4)\u3001\u5f0f(6)\u3001 \u5f0f(7)\u8207\u5f0f(8)\u52d5\u614b\u7522\u751f\u7684\u5404\u7a2e\u4e0d\u540c\u6982\u5ff5\u8a9e\u8a00\u6a21\u578b\u4ea6\u53ef\u518d\u900f\u904e\u7dda\u6027\u7d44\u5408\u65b9\u5f0f\u5206\u5225\u8207\u539f\u59cb N \u9023\u8a9e\u8a00\u6a21\u578b\u7d50\u5408\u4f86\u52d5\u614b\u8abf\u9069\u8a9e\u97f3\u8fa8\u8b58\u6240\u9700\u7684\u8a9e\u8a00\u6a21\u578b(\u5982\u5f0f(1)\u7684\u7d50\u5408\u65b9\u5f0f)\u3002\u5716 2", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "\u8a9e\u8a00\u6a21\u578b(Language Models, LM)\u5df2\u88ab\u5ee3\u6cdb\u5730\u4f7f\u7528\u65bc\u8a9e\u97f3\u8fa8\u8b58\u3001\u6a5f\u5668\u7ffb\u8b6f\u3001\u8cc7\u8a0a\u6aa2\u7d22\u4ee5\u53ca", | |
| "sec_num": null | |
| } | |
| ], | |
| "back_matter": [], | |
| "bib_entries": { | |
| "BIBREF0": { | |
| "ref_id": "b0", | |
| "title": "Modern Information Retrieval: the Concepts and Technology behind Search", | |
| "authors": [ | |
| { | |
| "first": "R", | |
| "middle": [], | |
| "last": "Baeza-Yates", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "B", | |
| "middle": [], | |
| "last": "Ribeiro-Neto", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2011, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Baeza-Yates, R., & Ribeiro-Neto, B. (2011). Modern Information Retrieval: the Concepts and Technology behind Search, Addison-Wesley Professional.", | |
| "links": null | |
| }, | |
| "BIBREF1": { | |
| "ref_id": "b1", | |
| "title": "Statistical language model adaptation: review and perspectives", | |
| "authors": [ | |
| { | |
| "first": "J", | |
| "middle": [ | |
| "R" | |
| ], | |
| "last": "Bellegarda", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2004, | |
| "venue": "Speech Communication", | |
| "volume": "42", | |
| "issue": "11", | |
| "pages": "93--108", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Bellegarda, J. R. (2004). Statistical language model adaptation: review and perspectives. Speech Communication, 42(11), 93-108.", | |
| "links": null | |
| }, | |
| "BIBREF2": { | |
| "ref_id": "b2", | |
| "title": "Build, compute, critique, repeat: Data analysis with latent variable models", | |
| "authors": [ | |
| { | |
| "first": "D", | |
| "middle": [ | |
| "M" | |
| ], | |
| "last": "Blei", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2014, | |
| "venue": "Annual Review of Statistics and Its Application", | |
| "volume": "1", | |
| "issue": "", | |
| "pages": "203--232", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Blei, D. M. (2014). Build, compute, critique, repeat: Data analysis with latent variable models. Annual Review of Statistics and Its Application, 1, 203-232.", | |
| "links": null | |
| }, | |
| "BIBREF3": { | |
| "ref_id": "b3", | |
| "title": "Topic models", | |
| "authors": [ | |
| { | |
| "first": "D", | |
| "middle": [ | |
| "M" | |
| ], | |
| "last": "Blei", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "J", | |
| "middle": [], | |
| "last": "Lafferty", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2009, | |
| "venue": "Text Mining: Theory and Applications", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Blei, D. M, & Lafferty, J. (2009). Topic models. in Srivastava, A., & Sahami, M., (eds.), Text Mining: Theory and Applications, Taylor and Francis, 2009.", | |
| "links": null | |
| }, | |
| "BIBREF4": { | |
| "ref_id": "b4", | |
| "title": "Latent Dirichlet Allocation", | |
| "authors": [ | |
| { | |
| "first": "D", | |
| "middle": [ | |
| "M" | |
| ], | |
| "last": "Blei", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "A", | |
| "middle": [ | |
| "Y" | |
| ], | |
| "last": "Ng", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "M", | |
| "middle": [ | |
| "I" | |
| ], | |
| "last": "Jordan", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2003, | |
| "venue": "Journal of Machine Learning Research", | |
| "volume": "3", | |
| "issue": "", | |
| "pages": "993--1022", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Blei, D. M., Ng, A. Y., & Jordan, M. I. (2003). Latent Dirichlet Allocation. Journal of Machine Learning Research, 3, 993-1022.", | |
| "links": null | |
| }, | |
| "BIBREF5": { | |
| "ref_id": "b5", | |
| "title": "Lightly supervised and data-driven approaches to Mandarin broadcast news transcription", | |
| "authors": [ | |
| { | |
| "first": "B", | |
| "middle": [], | |
| "last": "Chen", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "J.-W", | |
| "middle": [], | |
| "last": "Kuo", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "W.-H", | |
| "middle": [], | |
| "last": "Tsai", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2004, | |
| "venue": "Proceedings of the IEEE International Conference on Acoustics, Speech, Signal Processing", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "777--780", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Chen, B., Kuo, J.-W., & Tsai, W.-H. (2004). Lightly supervised and data-driven approaches to Mandarin broadcast news transcription. In Proceedings of the IEEE International Conference on Acoustics, Speech, Signal Processing, 777-780.", | |
| "links": null | |
| }, | |
| "BIBREF6": { | |
| "ref_id": "b6", | |
| "title": "Leveraging effective query modeling techniques for speech recognition and summarization", | |
| "authors": [ | |
| { | |
| "first": "K.-Y", | |
| "middle": [], | |
| "last": "Chen", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "S.-H", | |
| "middle": [], | |
| "last": "Liu", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "B", | |
| "middle": [], | |
| "last": "Chen", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "H.-M", | |
| "middle": [], | |
| "last": "Wang", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "W.-L", | |
| "middle": [], | |
| "last": "Hsu", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "H.-H", | |
| "middle": [], | |
| "last": "Chen", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "E.-E", | |
| "middle": [], | |
| "last": "Jan", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2014, | |
| "venue": "Proceedings of the Conference on Empirical Methods on Natural Language Processing", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Chen, K.-Y., Liu, S.-H., Chen, B., Wang, H.-M., Hsu, W.-L., Chen, H.-H., & Jan, E.-E. (2014). Leveraging effective query modeling techniques for speech recognition and summarization. In Proceedings of the Conference on Empirical Methods on Natural Language Processing.", | |
| "links": null | |
| }, | |
| "BIBREF7": { | |
| "ref_id": "b7", | |
| "title": "Maximum likelihood from incomplete data via the EM algorithm", | |
| "authors": [ | |
| { | |
| "first": "A", | |
| "middle": [ | |
| "P" | |
| ], | |
| "last": "Dempster", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "N", | |
| "middle": [ | |
| "M" | |
| ], | |
| "last": "Laird", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "D", | |
| "middle": [ | |
| "B" | |
| ], | |
| "last": "Rubin", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1977, | |
| "venue": "Journal of Royal Statistical Society B", | |
| "volume": "39", | |
| "issue": "1", | |
| "pages": "1--38", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Dempster, A. P., Laird , N. M., & Rubin, D. B. (1977). Maximum likelihood from incomplete data via the EM algorithm. Journal of Royal Statistical Society B, 39(1), 1-38.", | |
| "links": null | |
| }, | |
| "BIBREF8": { | |
| "ref_id": "b8", | |
| "title": "Deep Learning: Methods and Applications, Foundations and Trends in Signal Processing", | |
| "authors": [ | |
| { | |
| "first": "L", | |
| "middle": [], | |
| "last": "Deng", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "D", | |
| "middle": [], | |
| "last": "Yu", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2014, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Deng, L., & Yu, D. (2014). Deep Learning: Methods and Applications, Foundations and Trends in Signal Processing, Now Publishers.", | |
| "links": null | |
| }, | |
| "BIBREF9": { | |
| "ref_id": "b9", | |
| "title": "Topic-based language models using EM", | |
| "authors": [ | |
| { | |
| "first": "D", | |
| "middle": [], | |
| "last": "Gildea", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "T", | |
| "middle": [], | |
| "last": "Hofmann", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1999, | |
| "venue": "Proceedings of the European Conference on Speech Communication and Technology", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "2167--2170", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Gildea, D., & Hofmann, T. (1999). Topic-based language models using EM. In Proceedings of the European Conference on Speech Communication and Technology, 2167-2170.", | |
| "links": null | |
| }, | |
| "BIBREF10": { | |
| "ref_id": "b10", | |
| "title": "Fundamental technologies in modern speech recognition", | |
| "authors": [ | |
| { | |
| "first": "S", | |
| "middle": [], | |
| "last": "Furui", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "L", | |
| "middle": [], | |
| "last": "Deng", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "M", | |
| "middle": [], | |
| "last": "Gales", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "H", | |
| "middle": [], | |
| "last": "Ney", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "K", | |
| "middle": [], | |
| "last": "Tokuda", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2012, | |
| "venue": "IEEE Signal Processing Magazine", | |
| "volume": "29", | |
| "issue": "6", | |
| "pages": "16--17", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Furui, S., Deng, L., Gales, M., Ney, H., & Tokuda, K. (2012). Fundamental technologies in modern speech recognition. IEEE Signal Processing Magazine, 29(6), 16-17.", | |
| "links": null | |
| }, | |
| "BIBREF11": { | |
| "ref_id": "b11", | |
| "title": "Probabilistic latent semantic indexing", | |
| "authors": [ | |
| { | |
| "first": "T", | |
| "middle": [], | |
| "last": "Hofmann", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1999, | |
| "venue": "In Proceeding of the ACM Special Interest Group on Information Retrieval", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "50--57", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Hofmann, T. (1999). Probabilistic latent semantic indexing. In Proceeding of the ACM Special Interest Group on Information Retrieval, 50-57.", | |
| "links": null | |
| }, | |
| "BIBREF12": { | |
| "ref_id": "b12", | |
| "title": "A variational approximation for topic modeling of hierarchical corpora", | |
| "authors": [ | |
| { | |
| "first": "D", | |
| "middle": [], | |
| "last": "Kim", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "G", | |
| "middle": [ | |
| "M" | |
| ], | |
| "last": "Voelker", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "L", | |
| "middle": [ | |
| "K" | |
| ], | |
| "last": "Saul", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2013, | |
| "venue": "Proceedings of the International Conference on Machine Learning", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Kim, D.-k., Voelker, G. M., & Saul, L. K. (2013). A variational approximation for topic modeling of hierarchical corpora. In Proceedings of the International Conference on Machine Learning.", | |
| "links": null | |
| }, | |
| "BIBREF13": { | |
| "ref_id": "b13", | |
| "title": "Speech recognition and the frequency of recently used words: A modified Markov model for natural language", | |
| "authors": [ | |
| { | |
| "first": "R", | |
| "middle": [], | |
| "last": "Kuhn", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1988, | |
| "venue": "Proceedings of International Conference on Computational Linguistics", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "348--350", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Kuhn, R. (1988). Speech recognition and the frequency of recently used words: A modified Markov model for natural language. In Proceedings of International Conference on Computational Linguistics, 348-350.", | |
| "links": null | |
| }, | |
| "BIBREF14": { | |
| "ref_id": "b14", | |
| "title": "On information and sufficiency", | |
| "authors": [ | |
| { | |
| "first": "S", | |
| "middle": [], | |
| "last": "Kullback", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "R", | |
| "middle": [], | |
| "last": "Leibler", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1951, | |
| "venue": "Annals of Mathematical Statistics", | |
| "volume": "22", | |
| "issue": "1", | |
| "pages": "79--86", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Kullback, S., & Leibler, R. (1951). On information and sufficiency. Annals of Mathematical Statistics, 22(1), 79-86.", | |
| "links": null | |
| }, | |
| "BIBREF15": { | |
| "ref_id": "b15", | |
| "title": "Trigger-based language models: a maximum entropy approach", | |
| "authors": [ | |
| { | |
| "first": "R", | |
| "middle": [], | |
| "last": "Lau", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "R", | |
| "middle": [], | |
| "last": "Rosenfeld", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "S", | |
| "middle": [], | |
| "last": "Roukos", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1993, | |
| "venue": "Proceedings of the IEEE International Conference on Acoustics, Speech, Signal Processing", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "45--48", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Lau, R., Rosenfeld, R., & Roukos, S. (1993). Trigger-based language models: a maximum entropy approach. In Proceedings of the IEEE International Conference on Acoustics, Speech, Signal Processing, 45-48. \u4f7f\u7528\u6982\u5ff5\u8cc7\u8a0a\u65bc\u4e2d\u6587\u5927\u8a5e\u5f59\u9023\u7e8c\u8a9e\u97f3\u8fa8\u8b58\u4e4b\u7814\u7a76 59", | |
| "links": null | |
| }, | |
| "BIBREF16": { | |
| "ref_id": "b16", | |
| "title": "Training data selection for improving discriminative training of acoustic models", | |
| "authors": [ | |
| { | |
| "first": "S.-H", | |
| "middle": [], | |
| "last": "Liu", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "F.-H", | |
| "middle": [], | |
| "last": "Chu", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "S.-H", | |
| "middle": [], | |
| "last": "Lin", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "H.-S", | |
| "middle": [], | |
| "last": "Lee", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "B", | |
| "middle": [], | |
| "last": "Chen", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2007, | |
| "venue": "Proceedings of IEEE workshop on Automatic Speech Recognition and Understanding", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "284--289", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Liu, S.-H., Chu, F.-H., Lin, S.-H., Lee, H.-S., & Chen, B. (2007). Training data selection for improving discriminative training of acoustic models. In Proceedings of IEEE workshop on Automatic Speech Recognition and Understanding, 284-289.", | |
| "links": null | |
| }, | |
| "BIBREF17": { | |
| "ref_id": "b17", | |
| "title": "Recurrent neural network based language model", | |
| "authors": [ | |
| { | |
| "first": "T", | |
| "middle": [], | |
| "last": "Mikolov", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "M", | |
| "middle": [], | |
| "last": "Karafi\u00e1t", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "L", | |
| "middle": [], | |
| "last": "Burget", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "J", | |
| "middle": [], | |
| "last": "\u010cernock\u00fd", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "S", | |
| "middle": [], | |
| "last": "Khudanpur", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2010, | |
| "venue": "Proceedings of the Annual Conference of the International Speech Communication Association", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "1045--1048", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Mikolov, T., Karafi\u00e1t, M., Burget, L., \u010cernock\u00fd, J., & Khudanpur, S. (2010). Recurrent neural network based language model. In Proceedings of the Annual Conference of the International Speech Communication Association, 1045-1048.", | |
| "links": null | |
| }, | |
| "BIBREF18": { | |
| "ref_id": "b18", | |
| "title": "A word graph algorithm for large vocabulary continuous speech recognition", | |
| "authors": [ | |
| { | |
| "first": "S", | |
| "middle": [], | |
| "last": "Ortmanns", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "H", | |
| "middle": [], | |
| "last": "Ney", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "X", | |
| "middle": [], | |
| "last": "Aubert", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1997, | |
| "venue": "Computer Speech and Language", | |
| "volume": "11", | |
| "issue": "", | |
| "pages": "43--72", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Ortmanns, S., Ney, H., & Aubert, X. (1997). A word graph algorithm for large vocabulary continuous speech recognition. Computer Speech and Language, 11, 43-72.", | |
| "links": null | |
| }, | |
| "BIBREF19": { | |
| "ref_id": "b19", | |
| "title": "Speech information processing: Theory and applications", | |
| "authors": [ | |
| { | |
| "first": "D", | |
| "middle": [], | |
| "last": "O'shaughnessy", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "L", | |
| "middle": [], | |
| "last": "Deng", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "H", | |
| "middle": [], | |
| "last": "Li", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2013, | |
| "venue": "Proceedings of the IEEE", | |
| "volume": "101", | |
| "issue": "5", | |
| "pages": "1034--1037", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "O'Shaughnessy, D., Deng, L., & Li, H. (2013). Speech information processing: Theory and applications. Proceedings of the IEEE, 101(5), 1034-1037.", | |
| "links": null | |
| }, | |
| "BIBREF20": { | |
| "ref_id": "b20", | |
| "title": "Robust PLSA performs better than LDA", | |
| "authors": [ | |
| { | |
| "first": "A", | |
| "middle": [], | |
| "last": "Potapenko", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "V", | |
| "middle": [], | |
| "last": "Konstantin", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2013, | |
| "venue": "Proceedings of the European Conference on Information Retrieval", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "784--787", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Potapenko, A., & V. Konstantin. (2013). Robust PLSA performs better than LDA. In Proceedings of the European Conference on Information Retrieval, 784-787.", | |
| "links": null | |
| }, | |
| "BIBREF21": { | |
| "ref_id": "b21", | |
| "title": "Two decades of statistical language modeling: Where do we go from here? Proceedings of IEEE", | |
| "authors": [ | |
| { | |
| "first": "R", | |
| "middle": [], | |
| "last": "Rosenfeld", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2000, | |
| "venue": "", | |
| "volume": "88", | |
| "issue": "", | |
| "pages": "1270--1278", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Rosenfeld, R. (2000). Two decades of statistical language modeling: Where do we go from here? Proceedings of IEEE, 88(8), 2000, 1270-1278.", | |
| "links": null | |
| }, | |
| "BIBREF22": { | |
| "ref_id": "b22", | |
| "title": "SRI Language Modeling Toolkit. Available at", | |
| "authors": [ | |
| { | |
| "first": "A", | |
| "middle": [], | |
| "last": "Stolcke", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2000, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Stolcke, A. (2000). SRI Language Modeling Toolkit. Available at: http://www.speech.sri.com/projects/srilm/.", | |
| "links": null | |
| }, | |
| "BIBREF23": { | |
| "ref_id": "b23", | |
| "title": "Dynamic language model adaptation using variational Bayes inference", | |
| "authors": [ | |
| { | |
| "first": "Y", | |
| "middle": [], | |
| "last": "Tam", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "T", | |
| "middle": [], | |
| "last": "Schultz", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2005, | |
| "venue": "Proceedings of the Annual Conference of the International Speech Communication Association", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "5--8", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Tam, Y., & Schultz, T. (2005). Dynamic language model adaptation using variational Bayes inference. In Proceedings of the Annual Conference of the International Speech Communication Association, 5-8.", | |
| "links": null | |
| }, | |
| "BIBREF24": { | |
| "ref_id": "b24", | |
| "title": "Trigger-based language model adaptation for automatic meeting transcription", | |
| "authors": [ | |
| { | |
| "first": "C", | |
| "middle": [], | |
| "last": "Troncoso", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "T", | |
| "middle": [], | |
| "last": "Kawahara", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2005, | |
| "venue": "Proceedings of the Annual Conference of the International Speech Communication Association", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "1297--1300", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Troncoso, C., & Kawahara, T. (2005). Trigger-based language model adaptation for automatic meeting transcription. In Proceedings of the Annual Conference of the International Speech Communication Association, 1297-1300.", | |
| "links": null | |
| }, | |
| "BIBREF25": { | |
| "ref_id": "b25", | |
| "title": "MATBN: a Mandarin Chinese broadcast news corpus", | |
| "authors": [ | |
| { | |
| "first": "H.-M", | |
| "middle": [], | |
| "last": "Wang", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "B", | |
| "middle": [], | |
| "last": "Chen", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "J.-W", | |
| "middle": [], | |
| "last": "Kuo", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "S.-S", | |
| "middle": [], | |
| "last": "Cheng", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2005, | |
| "venue": "International Journal of Computational Linguistics & Chinese Language Processing", | |
| "volume": "10", | |
| "issue": "1", | |
| "pages": "219--235", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Wang, H.-M., Chen, B., Kuo, J.-W., & Cheng, S.-S. (2005). MATBN: a Mandarin Chinese broadcast news corpus. International Journal of Computational Linguistics & Chinese Language Processing, 10(1), 219-235.", | |
| "links": null | |
| }, | |
| "BIBREF26": { | |
| "ref_id": "b26", | |
| "title": "Statistical language models for information retrieval: A critical review", | |
| "authors": [ | |
| { | |
| "first": "C", | |
| "middle": [ | |
| "X" | |
| ], | |
| "last": "Zhai", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2008, | |
| "venue": "Foundations and Trends in Information Retrieval", | |
| "volume": "2", | |
| "issue": "3", | |
| "pages": "137--213", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Zhai, C. X. (2008). Statistical language models for information retrieval: A critical review. Foundations and Trends in Information Retrieval, 2(3), 137-213.", | |
| "links": null | |
| } | |
| }, | |
| "ref_entries": { | |
| "FIGREF0": { | |
| "uris": null, | |
| "type_str": "figure", | |
| "text": "\u3002\u6700\u91cd\u8981\u4e5f\u6700\u70ba\u5e38\u7528\u7684\u8a9e\u8a00\u6a21\u578b\u662f N \u9023(N-gram)\u8a9e\u8a00\u6a21 \u578b\uff0c\u8af8\u5982\u4e8c\u9023(Bigram)\u8207\u4e09\u9023(Trigram)\u8a9e\u8a00\u6a21\u578b\u3002N \u9023\u8a9e\u8a00\u6a21\u578b\u88ab\u7528\u4f86\u4f30\u6e2c\u6bcf\u4e00\u500b\u5f85\u9810\u6e2c \u8a5e\u5f59\u5728\u5176\u5148\u524d\u7dca\u9130\u7684 N-1 \u500b\u8a5e\u5f59\u5df2\u77e5\u7684\u60c5\u6cc1\u4e0b\u51fa\u73fe\u7684\u689d\u4ef6\u6a5f\u7387\uff1b\u7531\u6b64\u53ef\u77e5\uff0cN \u9023\u8a9e\u8a00\u6a21 \u578b\u662f\u5047\u8a2d\u6bcf\u4e00\u500b\u8a5e\u5f59\u51fa\u73fe\u7684\u6a5f\u7387\u50c5\u8207\u5b83\u7dca\u9130\u7684\u524d N-1 \u500b\u8a5e\u5f59\u6709\u95dc\uff0c\u4e26\u4ee5\u591a\u9805\u5f0f\u5206\u5e03 (Multinomial Distribution)\u8868\u793a\u4e4b\u3002\u7136\u800c N \u9023\u8a9e\u8a00\u6a21\u578b\u4ecd\u5b58\u5728\u8457\u8a31\u591a\u7f3a\u9ede\u9700\u8981\u6539\u5584\uff0c\u81f3\u5c11 \u6709\u4e09\u9ede\uff1a(1)N \u9023\u8a9e\u8a00\u6a21\u578b\u9650\u5236\u4e86 N \u7684\u5927\u5c0f\uff0c\u50c5\u80fd\u64f7\u53d6\u77ed\u8ddd\u96e2\u7684\u8a5e\u5f59\u898f\u5247\u8cc7\u8a0a\uff0c\u7121\u6cd5\u8003\u616e \u9577\u8ddd\u96e2\u7684\u8a9e\u53e5\u6216\u7bc7\u7ae0\u8cc7\u8a0a\uff1b(2)\u7576 N \u589e\u52a0\u6642\u4e0d\u50c5\u6703\u4f7f\u6a21\u578b\u53c3\u6578\u91cf\u5448\u73fe\u6307\u6578\u6027\u7684\u905e\u589e\uff0c\u9020\u6210 \u7a7a\u9593\u8207\u6642\u9593\u8907\u96dc\u5ea6\u5feb\u901f\u589e\u52a0\uff0c\u4e5f\u5bb9\u6613\u906d\u9047\u8cc7\u6599\u7a00\u758f\u3001\u7121\u6cd5\u70ba\u6bcf\u4e00\u7a2e\u8a5e\u5e8f\u5217\u7684\u6392\u5217\u7d44\u5408\u4f30 \u6e2c\u51fa\u6e96\u78ba\u7684\u6a5f\u7387\u503c\u7684\u554f\u984c\uff1b(3)N \u9023\u8a9e\u8a00\u6a21\u578b\u6975\u5bb9\u6613\u9762\u81e8\u8a13\u7df4\u8a9e\u6599\u8207\u6e2c\u8a66\u8a9e\u6599\u4e0d\u5339\u914d (Mismatch)\u800c\u9020\u6210\u7684\u4f30\u6e2c\u8aa4\u5dee\u3002\u6709\u9451\u65bc\u6b64\uff0c\u8fd1\u5341\u5e7e\u5e74\u4f86\u6709\u8a31\u591a\u52d5\u614b\u8a9e\u8a00\u6a21\u578b\u8abf\u9069\u6280\u8853\u88ab \u63d0\u51fa\uff0c\u7528\u4ee5\u767c\u5c55\u6709\u6548\u7684\u8a9e\u8a00\u6a21\u578b\u8f14\u52a9\u4e26\u5f4c\u88dc\u50b3\u7d71 N \u9023(N-gram)\u8a9e\u8a00\u6a21\u578b\u4e0d\u8db3\u4e4b\u8655\u3002\u5e38\u898b \u7684\u6709\u5feb\u53d6\u6a21\u578b(Cache Model)", | |
| "num": null | |
| }, | |
| "FIGREF1": { | |
| "uris": null, | |
| "type_str": "figure", | |
| "text": "\u7814 \u7a76 \u8a08 \u756b (MOST 103-2221-E-003-016-MY2, NSC 103-2911-I-003-301, NSC 101-2221-E-003-024-MY3 \u3001 NSC 101-2511-S-003-057-MY3 \u3001 NSC 101-2511-S-003-047-MY3 \u548c NSC 102-2221-E-003-014-MY3)\u4e4b\u7d93\u8cbb\u652f\u6301\uff0c\u8b39\u6b64\u81f4\u8b1d\u3002", | |
| "num": null | |
| }, | |
| "TABREF0": { | |
| "html": null, | |
| "text": "\u3002\u5728\u8a9e \u97f3\u8fa8\u8b58\u4efb\u52d9\u4e0a\uff0c\u5176\u4e3b\u8981\u7684\u529f\u80fd\u901a\u5e38\u662f\u85c9\u7531\u5df2\u89e3\u78bc\u7684\u6b77\u53f2\u8a5e\u5e8f\u5217(Word History)\u8cc7\u8a0a\u4f86\u9810\u6e2c \u4e0b\u4e00\u500b\u8a5e\u5f59(Upcoming Word)\u70ba\u4f55\u7684\u53ef\u80fd\u6027\u6700\u5927\uff0c\u4ee5\u5354\u52a9\u8a9e\u97f3\u8fa8\u8b58\u7cfb\u7d71\u5f9e\u773e\u591a\u6df7\u6dc6\u7684\u5019\u9078 \u8a5e\u5e8f\u5217\u5047\u8a2d(Candidate Word Sequence Hypotheses)\u4e2d\u627e\u51fa\u6700\u6709\u53ef\u80fd\u7684\u7d50\u679c", | |
| "content": "<table/>", | |
| "num": null, | |
| "type_str": "table" | |
| } | |
| } | |
| } | |
| } |