| { | |
| "paper_id": "O04-1021", | |
| "header": { | |
| "generated_with": "S2ORC 1.0.0", | |
| "date_generated": "2023-01-19T08:00:12.863323Z" | |
| }, | |
| "title": "", | |
| "authors": [], | |
| "year": "", | |
| "venue": null, | |
| "identifiers": {}, | |
| "abstract": "", | |
| "pdf_parse": { | |
| "paper_id": "O04-1021", | |
| "_pdf_hash": "", | |
| "abstract": [], | |
| "body_text": [ | |
| { | |
| "text": "EQUATION", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [ | |
| { | |
| "start": 0, | |
| "end": 8, | |
| "text": "EQUATION", | |
| "ref_id": "EQREF", | |
| "raw_str": "X w w w w w \u2212 = \uff0c\u8fa8\uf9fc\u8cc7\u8a0a\u5305\u542b\u6709 \u6b21\u97f3\u7bc0\u7684\u8a9e\u97f3\u65b7\u9ede\u8cc7\u8a0a\u3002\u6839\u64da\u6458\u8981\u6bd4\uf9b5\uff0c\u7cfb\u7d71\u6700\u5f8c\u53ef\u4ee5\u7372\u5f97\u9577\ufa01\u70ba N M Percentage = \u00d7 \u7684\u6458\u8981\u7d50\u679c \u3002 1 2 3 1 { , , ,..., , } N N Y w w w w w \u2212 = \u6458\u8981\uf9ca\u7a0b\u5982(\u5716 1)\u6240\u793a\uff0c\u5206\u6210\u4e0b\uf99c\u56db\u500b\u6b65\u9a5f\uff1a\u9996\u5148\u5c31\u8fa8\uf9fc\u7d50\u679c\u5c07 stop word \u53bb\u9664\uff0c\uf9b5\u5982\uff1a\u7684\u3001\u53ca\u3001\uf9ba\u7b49\uff0c \uf967\u5177\u8a9e\u7fa9\u8868\u793a\u7684\u8a5e\u3002\u518d\u8005\uff0c\u56e0\u70ba\uf967\u540c\u7684\u8a9e\u97f3\u6587\u4ef6\u53ef\u80fd\u5305\u542b\u7684\u91cd\u8981\u8cc7\u8a0a\uf97e\u4e26\uf967\u4e00\u81f4\uff0c\u6240\u4ee5\u6458\u8981\u58d3\u7e2e\u7684\u6bd4\uf9b5\u6703 \u5c0d\u6458\u8981\u7d50\u679c\u6709\u5f88\u5927\u7684\u5f71\u97ff\u3002\u56e0\u6b64\u9664\uf9ba\u53ef\u4ee5\u4f9d\u64da\u4f7f\u7528\u8005\u9700\u6c42\u8a2d\u5b9a\u6458\u8981\u6bd4\uf9b5\u5916\uff0c\u4e5f\u53ef\u4ee5\u85c9\u7531\u5224\u65b7\u5b57\u8a5e\u76f8\u5c0d\u65bc\u6587 \u7ae0\u6240\u4ee3\u8868\u7684\u91cd\u8981\u6027 \uff0c\u81ea\u52d5\u6c7a\u5b9a\u6458\u8981\u6bd4\uf9b5\u3002\u7b2c\u4e09\u6b65\u9a5f\uff0c\u5247\u662f\u5c07\u8a9e\u97f3\u8fa8\uf9fc\u4fe1\u8cf4\u5206\uf969\u3001\u91cd\u8981\u8a5e\u8a9e\u5206\uf969\u3001\u8a9e \u8a00\u5b78\u5206\uf969\u548c\u8a9e\u610f\u76f8\u4f9d\u5206\uf969\u7b49\u56db\u7a2e\u5206\uf969\u4f5c\u7d50\u5408\uff0c\u4ee5\u52d5\u614b\u898f\u5283\u641c\u5c0b\u7684\u65b9\u6cd5\uff0c\u5c0b\u627e\u53ef\u80fd\u7684\uf905\u63a5\u8a5e\u7d44\u3002 ( ) m R w 2 1 1 1 ( ) { ( ) ( ) ( | , ) ( , )} M C m R m L m m m B S D G m m m S Y C w R w L w w w B w w \u03bb \u03bb \u03bb \u03bb \u2212 \u2212 \u2212 = = + + + \u2211 (\u5f0f 1) \u5176\u4e2d\uff0c , , C R L \u03bb \u03bb \u03bb \u548c B \u03bb \u662f\u4ee3\u8868\u5404\u500b\u7279\u5fb5\uf96b\uf969\u7684\u6b0a\u91cd(weight)\uff0c\u7528\u4ee5\u7d50\u5408\u9019\u56db\u500b\u5206\uf969\u4e26\u4e14\u5e73\u8861\u5404\uf96b\uf969\u7684\u91cd\u8981\u6027\u3002 \u5716 1. \u81ea\u52d5\u8a9e\u97f3\u6587\u4ef6\u6458\u8981\u7a0b\u5e8f 2.1 \u8a9e\u97f3\u8fa8\uf9fc\u4fe1\u8cf4\u5206\uf969 \u8a9e\u97f3\u6458\u8981\u9700\u8981\u900f\u904e\u8fa8\uf9fc\u5668\u5f97\u5230\u8a9e\u8a00\u4e0a\u7684\u8cc7\u8a0a\uff0c\u4f46\u8a9e\u97f3\u8fa8\uf9fc\u53ef\u80fd\u6703\u7522\u751f\u8072\u5b78\u4e0a\u548c\u8a9e\u8a00\u5b78\u4e0a\u7684\u8fa8\uf9fc\u932f\u8aa4\uff0c \u64fe\uf91b\u6700\u5f8c\u6458\u8981\u7d50\u679c\u7684\u610f\u7fa9\u3002\u56e0\u6b64\uff0c\u6211\u5011\u5c07\u8a9e\u97f3\u8fa8\uf9fc\u7684\u4fe1\u8cf4\u5206\uf969 \u5f15\u5165\uff0c\u76ee\u7684\u5728\u65bc\u9078\u64c7\u8fa8\uf9fc\u8f03\u6b63\u78ba\u7684\u7d50 \u679c\uff0c\u4f5c\u70ba\u5224\u65b7\u9078\u64c7\u6458\u8981\u55ae\u5143\u7684\u5206\uf969\u4e4b\u3127\u3002\u7d71\u8a08\u5f0f\u8a9e\u97f3\u8fa8\uf9fc\u662f\u57fa\u65bc\u8c9d\u6c0f\u6cd5\u5247\uff0c\u4fe1\u8cf4\u5206\uf969\u662f\u4f30\u7b97\u8a9e\u97f3\u8fa8\uf9fc\u4e2d\uff0c \u7ed9 \u5b9a \u4e00 \uf905 \u89c0 \u6e2c \u8a9e \u97f3 \u5e8f \uf99c ( ) m C w 1 ,..., t t x x = x m \u5c0d \u65bc \u4e00 \u5b57 \uf905 1 ,..., m w w w = \uff0c \u8a08 \u7b97 \u5176 \u4e8b \u5f8c \u6a5f \uf961 (posterior probability) \u3002\u8fa8\uf9fc\u7684\u968e\u6bb5\u4e2d\uff0c\u6211\u5011\u671f\u671b\u80fd\u5920\u5f97\u5230\u6700\u5927\u7684\u4e8b\u5f8c\u6a5f\uf961\u503c\uff0c\u4e5f\u5c31\u662f\u80fd\u5920\u6709\u8f03\u5c0f\u7684\u8aa4\u5dee\uff0c \u6240\u4ee5\u53ef\u5f97\u4e0b\uf99c\u5f0f\u5b50\uff1a ( | m t p w x ) ( ) max ( | ) max ( | ) ( ) ( ) max ( | ) ( m m t t m m t t m C w p w x p x w p w p x p x w p w = = \u22c5 = \u22c5 ) m ) (\u5f0f 2) \u5176\u4e2d\uff0c ( m p w \u70ba\u8a9e\u8a00\u6a21\u578b\u7684\u6a5f\uf961\u3002 \u70ba\u8072\u5b78\u6a21\u578b\u7684\u6a5f\uf961\u3002 \u70ba\u89c0\u6e2c\u5230\u8072\u5b78\u7279\u5fb5\u7684\u6a5f\uf961\u3002 ( | ) t m p x w ( ) t p x 2.2 \u91cd\u8981\u8a5e\u5206\uf969 \u8981\u5c0d\u8fa8\uf9fc\u7d50\u679c\u505a\u8a9e\u97f3\u6587\u4ef6\u6458\u8981\u7684\u8655\uf9e4\u6642\uff0c\u9996\u5148\u9700\u8981\u5c07\u8a9e\u97f3\u6587\u4ef6\u5167\u5c6c\u65bc\u91cd\u8981\u7684\u8a5e\u8a9e\u4fdd\uf9cd\u4e0b\uf92d\uff0c\u800c\u628a\uf967\u5177 \u5099\u6709\u8868\u9054\u6587\u7ae0\u610f\u7fa9\u7684\u8a5e\u8207\u5b57\u62bd\uf9ea\u3002\u6211\u5011\u5f15\u7528\u4e00\u7d44\u6a19\u984c\u672c\u6587\u4e92\u76f8\u5c0d\u7167\u7684\u65b0\u805e\u8a9e\uf9be\u5eab\uff0c\uf92d\u8f14\u52a9\u5224\u65b7\u8fa8\uf9fc\u7d50\u679c\u7684 \u8a5e\uf906\u662f\u5426\u5177\u6709\u4ee3\u8868\u6027\u3002\u5be6\u9a57\u5f9e\u516c\u5171\u96fb\u8996\u65b0\u805e\u6536\u96c6 2001 \u5230 2002 \uf98e\u7684\u65b0\u805e\uff0c\u6574\uf9e4\uf978\u5343\uf9b2\uf9d1\u5247\u7684\u65b0\u805e\u5831\u5c0e\u8a9e\uf9be\u3002 \u70ba\uf9ba\u6aa2\uf96a\u51fa\u8207\u6458\u8981\u6587\u7ae0\u5167\u5bb9\u76f8\u4f3c\u7684\u65b0\u805e\u5831\u5c0e\u8a9e\uf9be\uff0c\u6211\u5011\uf96b\u7167\u8cc7\u8a0a\u6aa2\uf96a\u7684\u6280\u8853(Information Retrieval, IR) [1]\uff0c \u9996\u5148\u5c07\u5e73\ufa08\u8a9e\uf9be\u7684\u6240\u6709\u6587\u7ae0\u5167\u5bb9\uff0c\u5206\u5225\u8f49\u63db\u6210\u4ee5\u8a5e \u548c\u97f3\u7bc0 w d v s d v \u70ba\u55ae\u5143\u7684\uf978\u500b\u5411\uf97e\uff0c\u5c0d\u65bc\u6240\u8981\u6458\u8981\u7684\u8a9e\u97f3 \u6587\u4ef6\u4e5f\u540c\u6a23\u5730\u505a\u8f49\u63db\u70ba\uf978\u500b\u5411\uf97e\uff0c\u53ef\u8868\u793a\u6210 \u548c 1 2 ( , ,..., ) P w w w w d d d d v t t t = 1 2 ( , ,..., ) Q s s s s d d d d v t t t = \u3002\u5176\u4e2d\uff0c Q \u8868\u793a\u4ee5\u97f3 \u7bc0\u55ae\u5143\u70ba\u57fa\u790e\u7684\u5411\uf97e s d v \u7dad\ufa01\uff0c\u4f9d\u64da\u56db\u767e\uf9b2\u4e8c\u500b\u4e2d\u6587\u97f3\u7bc0\uff0c\u4e26\u8003\u616e\u8a5e\u9577\u70ba\u4e8c\u7684\u6240\u6709\u914d\u5c0d\u7d44\u5408\uff0c\u7522\u751f\u7dad\ufa01\u70ba \u7684\u5411\uf97e\u3002\u800c \u5247\u8868\u793a\u4ee5\u8a5e\u70ba\u57fa\u790e\u7684\u5411\uf97e \u7dad\ufa01\uff0c\u6839\u64da\u8fad\u5178\u5167\u6240\u5b9a\u7fa9\u7684\u8a5e\uff0c\uf967 \u8003\u616e\u865b\u8a5e(stop word)\u7684\u90e8\u5206\uff0c\u56e0\u70ba\u865b\u8a5e\uf967\u6703\u5f71\u97ff\u6587\u7ae0\u5167\u5bb9\u610f\u7fa9\u7684\u6aa2\uf96a\uff0c\u53bb\u9664\u7528\u4ee5\ufa09\u4f4e\u8a08\u7b97\u7684\u7dad\ufa01\uff0c\u5f97\u5230\u7d50 \u679c (402 402 402) 162006 Q = + \u00d7 = P w d v 28000 P = \u3002\uf978\u5411\uf97e\u5167\u7684\u4e4b\uf969\u503c\u4ee5\u8a5e\u983b\u548c\u53cd\u8f49\u6587\u4ef6\u983b\uf961\u8868\u793a(term frequency * inverse document frequency, tf.idf)[8]\u3002\u540c\u6642\uff0c\u5fc5\u9808\u8003\u616e\u8a9e\u97f3\u8fa8\uf9fc ( ) j C w \u53ef\u80fd\u9020\u6210\u7684\u5f71\u97ff\uff0c\u5c07\u8fa8\uf9fc\uf967\u597d\u7684\u7d50\u679c\uff0c\u6e1b\u4f4e\u5206\uf969\u3002\u56e0\u6b64\u6bcf\u4e00\u500b\uf96a \u5f15\u503c\u7684\u8a08\u7b97\u65b9\u6cd5\u5982\u4e0b\uff1a ( ) ln( 1) ln( /( 1) j j j w d j w w t C w f N df = \u22c5 + \u22c5 + ) (\u5f0f 3) \u7d50\u5408\uf978\u5411\uf97e\uf92d\u505a\u6587\u4ef6\u67e5\u8a62\uff0c\uf9dd\u7528\u5411\uf97e\u5167\u7a4d\u7684\u8a08\u7b97\uff0c\u67e5\u8a62\u5e73\ufa08\u8a9e\uf9be\u5167\u6240\u6709\u6587\u7ae0\u7684\u95dc\uf997 \uff0c ( , ) R q d 2 2 ( , ) cos( , ) (1 )cos( , ) ( ) /( ) (1 )( ) /( ) w w w w s s s s R q d R q d w w wT w w w w s s sT s s s s R q d q d R q d q d R q d v S v S v S v S v S v v S v S v S v v S v S \u03b1 \u03b1 \u03b1 \u03b1 = + \u2212 = \u22c5 \u22c5 + \u2212 \u22c5 (\u5f0f 4) \u4e26\u4e14\u61c9\u7528\uf96b\uf969 0.2 R \u03b1 = \uf92d\u5e73\u8861\u5b57\u8207\u97f3\u7bc0\uf978\u500b\u5411\uf97e\u7684\u6b0a\u91cd\u3002\u4f9d\u64da\u6b64\u95dc\uf997\u5206\uf969 \uff0c\u627e\u51fa\u4e00\u7bc7\u6587\u4ef6\u63cf\u8ff0\u7684\u65b0 \u805e\u4e8b\u4ef6\u6700\u63a5\u8fd1\u7684\u6587\u7ae0 \uff0c\u4e4b\u5f8c\uff0c\u4ee5\u6f5b\u85cf\u5f0f\u8a9e\u610f\u5206\u6790\uf96a\u5f15\u4f7f\u7528\u5411\uf97e\u7a7a\u9593\u7684\u65b9\u6cd5[8]\uff0c\u641c\u5c0b\u8fa8 \uf9fc\uf906\u5b50\u7684\u8a5e\u8207\u5e73\ufa08\u8a9e\uf9be\u6a19\u984c\u5167\u7684\u8a5e\uff0c\uf978\u8005\u4e4b\u9593\u5b58\u5728\u7684\u95dc\u4fc2\u3002 ( , ) R q d * arg max ( , ) d d R = qd \u65b9\u6cd5\uf96f\u660e\u5982(\u5716 2)\u6240\u793a\uff0c\u9996\u5148\u6839\u64da\u5e73\ufa08\u8a9e\uf9be\u548c\u8fad\u5178\uff0c\u5efa\uf9f7\u4e00\u500b\u6587\u7ae0\u53ca\u8a5e\u7684\u4e8c\u7dad\u77e9\u9663 w t d A \u00d7 \uff0c\u7dad\ufa01\u70ba \u3002\u7d93\u7531\u8a5e\u5c0d\u61c9\u65bc\u6587\u7ae0\u4ee5\u53ca\u6587\u7ae0\u5c0d\u61c9\u8a5e\u7684\u95dc\u4fc2 ( ) 2006 5104 \u00d7 ( ) terms documents documents terms \u00d7 \u22c5 \u00d7 \uff0c\u6700\u5f8c\u53ef \u4ee5\u63a8\u5c0e\u51fa\u8a5e\u5c0d\u8a5e\u7684\u95dc\uf997 T AA terms terms = \u00d7 \u3002\u914d\u5408\u5947\uf962\u503c\u5206\u89e3\u65b9\u6cd5\uf92d\u9054\u5230\u7dad\ufa01\u7684\ufa09\u4f4e\uff0c\u5c07\u5171\u540c\u767c\u751f\u7684\u4e8b\u4ef6 \u6295\u5f71\u5230\u76f8\u540c\u7684\u7dad\ufa01\u4e0a\u3002\u900f\u904e\u5947\uf962\u503c\u5206\u89e3 \uff0c\u5c07\u77e9\u9663\u5206\u89e3\u6210\u4e09\u500b\u77e9\u9663 \uff0c \u548c ( \uff0c \u5176\u4e2d \u3002 ( ) T t d t n n n d n A U S V \u00d7 \u00d7 \u00d7 \u00d7 = t k U \u00d7 k k S \u00d7 ) T d k V \u00d7 min( , ) n t = d \u5716 2. \u5947\uf962\u503c\u5206\u89e3 \u5c07\u53d6\u5c0d\u89d2\u77e9\u9663\uf94f\u8a08\u8b8a\uf962\uf97e\u4e4b\u767e\u5206\u4e4b\u4e5d\u5341\u4f5c\u70ba\u7dad\ufa01\ufa09\u4f4e\u7684\u4f9d\u64da k n < \u3002\u77e9\u9663\u4e2d\u6bcf\u4e00\u500b\u6210\u5206\u7684\u503c\uff0c\u7528 tf idf \u00d7 \u4ee3 \u8868\u3002\u8a5e\u5c0d\u8a5e\u7684\u77e9\u9663\u900f\u904e\ufa09\u7dad\ufa01\u7684\u8cc7\u8a0a\uff0c\uf92d\u8a08\u7b97 2 T T AA US V = \u3002\u900f\u904e\u65b0\u7684\u8a5e\u5c0d\u8a5e\u77e9\u9663\uf965\u53ef\u4ee5\u5f97\u77e5\uf978\u500b\u8a5e\u7684\u76f8 \u4f3c\u6027 ( , ) LSI i j P w w \uff0c\u5176\u5206\uf969\u8a08\u7b97\u65b9\u6cd5\u5982\u4e0b\uff1a 2 ( , ) cos( , ) w w w w w w wT w w w w LSI i j i j i j i j P w w U S U S U S U U S U S = = i (\u5f0f 5) \u6700\u5f8c\u6b78\u7d0d\u4e0a\u8ff0\u7684\u6b65\u9a5f\uff0c\u900f\u904e\u4e0b\uf99c\u7a0b\u5e8f\u7684\u8a08\u7b97\u65b9\u6cd5\uff0c\u6211\u5011\u53ef\u4ee5\u5f9e\u5e73\ufa08\u8a9e\uf9be\u4e2d\u8403\u53d6\u91cd\u8981\u7684\u8cc7\u8a0a ( ) i R w \u3002\u5176\u4e2d * t j w \u4ee3\u8868\u5e73\ufa08\u8a9e\uf9be\u6a19\u984c\u5167\u7684\u8a5e\uff0c\u800c \u662f\u8f38\u5165\u6587\u7ae0\u7d93\u8a9e\u97f3\uf9fc\u8fa8\u5f8c\u7684\u8a5e\uff0c\u56e0\u6b64\u53ef\u8a08\u7b97\u51fa \u5c0d\u65bc\u6458\u8981\u6587\u4ef6\u7684\u91cd\u8981\u6027\uff1a i w i w * ( ) max{ ( , ) ( 1) ln( /( 1))} i i t i L S I i j w w j R w P w w f N df = \u22c5 + \u22c5 + (", | |
| "eq_num": "\u5f0f 6" | |
| } | |
| ], | |
| "section": "", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "( | , ) ( , , ) ( , ) ( , ) ( ) ( ) ( ) i L w w w p F w w w F w w p F w w F w p F w F w = \u22c5 + \u22c5 + \u22c5 \u2211 (\u5f0f 7)", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "\u5176\u4e2d\uff0c \u8868\u793a\u6b63\uf969\u7684\u6b0a\u91cd\u4e14\u5408\u70ba\u4e00\u3002 ", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "EQUATION", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [ | |
| { | |
| "start": 0, | |
| "end": 8, | |
| "text": "EQUATION", | |
| "ref_id": "EQREF", | |
| "raw_str": "( | ) ( , | ) ( | ) ( | , P t s P B D s P B s P D s B = = \u00d7 ) (\u5f0f 8) \u5047\u5b9a\u6bcf\u4e00\u500b\u76f8\u4f9d\u95dc\u4fc2\ufa26\u662f\u7368\uf9f7\u7684\uff0c\u4e14\u5256\u6790\u5f8c\u6bcf\u4e00\u500b\u8a5e \uff0c\ufa26\u76f8\u4f9d\u65bc\u67d0\u4e00\u500b\u4e2d\u5fc3\u8a5e \uff0c\u5176\u76f8\u4f9d\u7684\u95dc\uf997\u53ef\u4ee5 \u754c\u5b9a\u70ba \u3002\u56e0\u6b64\uff0c\u76f8\u4f9d\u95dc\uf997\u53ef\u4ee5\u91cd\u65b0\u5b9a\u7fa9\u6210\u4e00\u500b\u96c6\u5408 \uff0c\u8868\u793a\u5982\u4e0b\uff1a m w m w h m w m w h R , { ( , , )} i i w i i w w h d w h R , 1 ( | , ) ( ( , , )) j j w j n j w w h j P D s B P d w h R = = \u220f (\u5f0f 9) \u5728\u8a08\u7b97\uf978\u500b\u8a5e \u548c i w j w \uff0c\u5b58\u5728\u76f8\u4f9d\u95dc\u4fc2 \u7684\u6a5f\uf961 R ( | , ) i j F R w w \u6642\uff0c\u540c\u4e00\u95dc\u4fc2\u53ef\u8868\u793a\u5982\u4e0b\uff1a ( | , ) ( , , ) ( , i j i j i j ) F R w w C R w w C w w = (\u5f0f 10) \u5176\u4e2d\uff0c \u8868\u793a\u70ba\uf978\u500b\u8a5e\u4e00\u8d77\u51fa\u73fe\u7684\u983b\uf961\uff0c \u8868\u793a\uf978\u500b\u8a5e\u4e00\u8d77\u51fa\u73fe\u6642\u5b58\u5728\u6709\u7684\u76f8\u4f9d\u95dc\u4fc2\u3002 \u4e14\u70ba\uf9ba\u907f\u514d\u8cc7\uf9be\u7a00\u758f(sparse data)\u7684\u554f\u984c\uff0c\u9032\u4e00\u6b65\u5730\uf9dd\u7528\u77e5\u7db2\u7684\u77e5\uf9fc\uff0c\u5c07\u8a5e\u8f49\u63db\u6210\u76f8\u5c0d\u61c9\u7684\u4e0a\u4f4d\u8a5e (hypernym)\uff0c\u4ee5\u8868\u793a\u4e4b ( , ) i j C w w ( , , ) i j C R w w ( ) H i \uff0c\u5f97\u5230\u4e0b\uf99c\u5f0f\u5b50\uff1a ( | ( ), ( )) ( , ( ), ( )) ( ( ), ( )) i j i j i j F R H w H w C R H w H w C H w H w = (", | |
| "eq_num": "\u5f0f 11" | |
| } | |
| ], | |
| "section": "", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "EQUATION", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [ | |
| { | |
| "start": 0, | |
| "end": 8, | |
| "text": "EQUATION", | |
| "ref_id": "EQREF", | |
| "raw_str": ") B w w f T S w w f S w w N = = \u2211\u2211\u2211 \u00d7 (\u5f0f 12) \u5176\u4e2d\uff0c i T f \u8868\u793a\u6587\u6cd5\u5256\u6790 PCFG \u4e4b\u6a5f\uf961\u3002 k i DR f \u8868\u793a\u8a9e\u610f\u76f8\u95dc\u6cd5\u5247\u4e4b\u6a5f\uf961 s N \u8868\u793a\uf906\u5b50\u7e3d\uf969\u3002 \u8868 \u793a \u4e00\u500b\uf906\u5b50\u5305\u542b\u6709 \u548c \u3002 \u8868 \u793a \u91dd \u5c0d \uf906 \u5b50 \u53ef \u80fd \u5256 \u6790 \u7684 \u6587 \u6cd5 \u6a39 \u3002 k \u8868 \u793a \u5b58 \u5728 \u7684 \u95dc \uf99a \u6027 \uf96a \u5f15 \u3002 \u6307\u9577\ufa01 \u7684\uf906\u5b50\u5b58\u5728\u76f8\u4f9d\u95dc\u4fc2\u3002 \u8003\u616e\u8a13\uf996\u8a9e\uf9be\u7a00\u758f\u7684\u554f\u984c(sparse data)\uff0c \u56e0\u6b64\u4f7f HowNet \u5167\u5b9a\u7fa9\u7684\u4e0a\u4f4d\u8a5e(Hypernym)\u53d6\u4ee3\u539f\u672c\u7684\u8a5e\u7d44\uff1a j S a w b w i T j S { ( , )|1 1 k i i a b w D DR w w k N = \u2264 \u2264 } \u2212 j w N ( , )", | |
| "eq_num": "( ( )" | |
| } | |
| ], | |
| "section": "", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "i j i j SC w w SC w SC w = \u2212 ; 1 1 1 1 ( ) (( [ ] )/( [ ])) F N N i t t n n SC w M n n M n F = = = = \u00d7 \u00d7 \u2211 \u2211 \u2211 t (\u5f0f 16) \u5176\u4e2d\uff0c [ ] t M n \u5085\uf9f7\uf96e\u8f49\u63db\u5f37\ufa01\uff1b \u983b\u6846\uf96a\u5f15\uff1b \u97f3\u8a0a\u5206\u9801\uf96a\u5f15\u3002 n t 2) \u983b\u8b5c\uf904\u52d5\uff0c\u540c\u6a23\u8868\u793a\u983b\u8b5c\u4e0a\u7279\u5fb5\uff0c\u53ef\u6e2c\uf97e\uf978\u55ae\u5143\u9593\u7684\u5dee\uf962\uff0c 1 1 1 ( ) (0.85 [ ]) F N i t t n SR w M n F = = = \u00d7 \u2211 \u2211 \u3002 3) \u983b \u8b5c \u8b8a \u9077 \uff0c \u6b63 \u898f \u5283 \u76f8 \u9130 \u983b \u8b5c \u7684 \u5e73 \u65b9 \u5dee \uff0c \u76ee \u7684 \u5728 \u65bc \uf97e \u6e2c \u983b \u8b5c \u4e0a \u7684 \u5c40 \u90e8 \u8b8a \u5316 \uff0c 2 1 1 1 1 ( ) ( [ ] [ ]) F N i t t t n SF w N n N n F \u2212 = = = \u2212 \u2211\u2211 \u3002\u5176\u4e2d\uff0c \u5b9a\u7fa9\u5728\u7b2c t \u97f3\u8a0a\u5206\u9801\u7684\u6b63\u898f\u5316\u5085\uf9f7\uf96e\u5f37\ufa01\u3002 [ ] t N n 4) \u6642 \u57df \u4e0a \u8d8a \uf9b2 \uf961 \uff0c \u4e00 \u822c \u7528 \u65bc \u566a \u97f3 \u5075 \u6e2c \uff0c \u5728 \u6b64 \u53ef \u77e5 \uf978 \u55ae \u5143 \u9593 \uff0c \u566a \u97f3 \u6539 \u8b8a \u7a0b \ufa01 \u3002 1 1 1 1 ( ) ( ( [ ]) ( [ 1])) 2 F N i t t n ZCR w sign s n sign s n F = = = \u2212 \u2211 \u2211 t \u2212 \u3002 5) \u6885\u723e\u5012\u983b\u8b5c\uf96b\uf969\uff0c\u61c9\u7528\u8a9e\u97f3\u8fa8\uf9fc\u5e38\u7528\u7684\u6885\u723e\u5012\u983b\u8b5c\uf96b\uf969\uff0c\u5171\u53d6\u4e09\u5341\u4e5d\u7dad\uff0c\u4e3b\u8981\u662f\u6a21\u64ec\u4eba\u7684\u807d\u89ba\u6a21\u578b\uff0c", | |
| "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": "1 1 mAP q i N N i k q i k N N rank = = = \u2211 \u2211 i k \uff1b 1 1 rAP q N i i q N N N = = \u2211 (", | |
| "eq_num": "\u5f0f 18" | |
| } | |
| ], | |
| "section": "", | |
| "sec_num": null | |
| } | |
| ], | |
| "back_matter": [], | |
| "bib_entries": { | |
| "BIBREF0": { | |
| "ref_id": "b0", | |
| "title": "Discriminating Capabilities of Syllable-Based Features and Approaches of Utilizing Them for Voice Retrieval of Speech Information in Mandarin Chinese", | |
| "authors": [ | |
| { | |
| "first": "Hsin-Min", | |
| "middle": [], | |
| "last": "Berlin Chen", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "", | |
| "middle": [], | |
| "last": "Wang", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Ieee", | |
| "middle": [], | |
| "last": "Member", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Lin-Shan", | |
| "middle": [], | |
| "last": "Lee", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2002, | |
| "venue": "IEEE TRANSACTIONS ON SPEECH AND AUDIO PROCESSING", | |
| "volume": "10", | |
| "issue": "5", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Berlin Chen, Hsin-min Wang, Member, IEEE, and Lin-shan Lee, Fellow, IEEE, \"Discriminating Capabilities of Syllable-Based Features and Approaches of Utilizing Them for Voice Retrieval of Speech Information in Mandarin Chinese,\" IEEE TRANSACTIONS ON SPEECH AND AUDIO PROCESSING, VOL. 10, NO. 5, JULY 2002", | |
| "links": null | |
| }, | |
| "BIBREF1": { | |
| "ref_id": "b1", | |
| "title": "A Trainable Document Summarizer", | |
| "authors": [ | |
| { | |
| "first": "Julian", | |
| "middle": [], | |
| "last": "Kupiec", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Jan", | |
| "middle": [], | |
| "last": "Pedersen", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Francine", | |
| "middle": [], | |
| "last": "Chen", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": null, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Julian Kupiec, Jan Pedersen and Francine Chen, \"A Trainable Document Summarizer\", Xerox Palo Alto Research Center", | |
| "links": null | |
| }, | |
| "BIBREF2": { | |
| "ref_id": "b2", | |
| "title": "NEWSCAST SPEECH SUMMARIZATION VIA SENTENCE SHORTENING BASED ON PROSODIC FEATURES", | |
| "authors": [ | |
| { | |
| "first": "Kiyonori", | |
| "middle": [], | |
| "last": "Ohtake", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Kazuhide", | |
| "middle": [], | |
| "last": "Yamamoto", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Yuji", | |
| "middle": [], | |
| "last": "Toma", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Shiro", | |
| "middle": [], | |
| "last": "Sado", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Shigeru", | |
| "middle": [], | |
| "last": "Masuyama", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Seiichi", | |
| "middle": [], | |
| "last": "Nakagawa", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": null, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Kiyonori Ohtake, Kazuhide Yamamoto, Yuji Toma, Shiro Sado, Shigeru Masuyama,and Seiichi Nakagawa, \"NEWSCAST SPEECH SUMMARIZATION VIA SENTENCE SHORTENING BASED ON PROSODIC FEATURES\", Toyohashi University of Technology, Japan", | |
| "links": null | |
| }, | |
| "BIBREF3": { | |
| "ref_id": "b3", | |
| "title": "A New Approach to Automatic Speech Summarization", | |
| "authors": [ | |
| { | |
| "first": "Chiori", | |
| "middle": [], | |
| "last": "Hori", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Ieee", | |
| "middle": [], | |
| "last": "Member", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Sadaoki", | |
| "middle": [], | |
| "last": "Furui", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "", | |
| "middle": [], | |
| "last": "Fellow", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "", | |
| "middle": [], | |
| "last": "Ieee", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2003, | |
| "venue": "IEEE TRANSACTIONS ON MULTIMEDIA", | |
| "volume": "5", | |
| "issue": "3", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Chiori Hori, Member, IEEE, and Sadaoki Furui, Fellow, IEEE, \"A New Approach to Automatic Speech Summarization,\" IEEE TRANSACTIONS ON MULTIMEDIA, VOL. 5, NO. 3, SEPTEMBER 2003", | |
| "links": null | |
| }, | |
| "BIBREF4": { | |
| "ref_id": "b4", | |
| "title": "Speech-to-Text and Speech-to-Speech Summarization of Spontaneous Speech", | |
| "authors": [ | |
| { | |
| "first": "S", | |
| "middle": [], | |
| "last": "Furui", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "T", | |
| "middle": [], | |
| "last": "Kikuchi", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Y", | |
| "middle": [], | |
| "last": "Shinnaka", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "C", | |
| "middle": [], | |
| "last": "Hori", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2004, | |
| "venue": "Speech and Audio Processing", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "401--408", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Furui, S.; Kikuchi, T.; Shinnaka, Y.; Hori, C., \"Speech-to-Text and Speech-to-Speech Summarization of Spontaneous Speech,\" Speech and Audio Processing, IEEE Transactions on , Volume: 12 , Issue: 4 , July 2004, pp. 401 -408", | |
| "links": null | |
| }, | |
| "BIBREF5": { | |
| "ref_id": "b5", | |
| "title": "Musical genre classification of audio signals", | |
| "authors": [ | |
| { | |
| "first": "G", | |
| "middle": [], | |
| "last": "Tzanetakis", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "P", | |
| "middle": [], | |
| "last": "Cook", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2002, | |
| "venue": "IEEE Transactions on Speech and Audio Processing", | |
| "volume": "10", | |
| "issue": "5", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "G. Tzanetakis and P. Cook, \"Musical genre classification of audio signals,\" IEEE Transactions on Speech and Audio Processing, vol. 10, No. 5, July 2002.", | |
| "links": null | |
| }, | |
| "BIBREF6": { | |
| "ref_id": "b6", | |
| "title": "Minimum Classification Error Rate Methods for Speech Recognition", | |
| "authors": [ | |
| { | |
| "first": "", | |
| "middle": [], | |
| "last": "Biing-Hwang Juang", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "", | |
| "middle": [], | |
| "last": "Fellow", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Wu", | |
| "middle": [], | |
| "last": "Ieee", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "", | |
| "middle": [], | |
| "last": "Chou", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Ieee", | |
| "middle": [], | |
| "last": "Member", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Chin-Hui", | |
| "middle": [], | |
| "last": "Lee", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "", | |
| "middle": [], | |
| "last": "Fellow", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "", | |
| "middle": [], | |
| "last": "Ieee", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1997, | |
| "venue": "IEEE TRANSACTIONS ON SPEECH AND AUDIO PROCESSING", | |
| "volume": "5", | |
| "issue": "3", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Biing-Hwang Juang, Fellow, IEEE, Wu Chou, Member, IEEE, and Chin-Hui Lee, Fellow, IEEE, \"Minimum Classification Error Rate Methods for Speech Recognition,\" IEEE TRANSACTIONS ON SPEECH AND AUDIO PROCESSING, VOL. 5, NO. 3, MAY 1997", | |
| "links": null | |
| }, | |
| "BIBREF7": { | |
| "ref_id": "b7", | |
| "title": "Foundations of Statistical Natural Language Processing", | |
| "authors": [ | |
| { | |
| "first": "D", | |
| "middle": [], | |
| "last": "Christopher", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Hinrich", | |
| "middle": [], | |
| "last": "Manning", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "", | |
| "middle": [], | |
| "last": "Schutze", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1999, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Christopher D. Manning and Hinrich Schutze, \"Foundations of Statistical Natural Language Processing\", The MIT Press, 1999", | |
| "links": null | |
| }, | |
| "BIBREF8": { | |
| "ref_id": "b8", | |
| "title": "Variable N-Grams and Extensions for Conversational Speech LanguageModeling", | |
| "authors": [ | |
| { | |
| "first": "Manhung", | |
| "middle": [], | |
| "last": "Siu", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Ieee", | |
| "middle": [], | |
| "last": "Member", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Mari", | |
| "middle": [], | |
| "last": "Ostendorf", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2000, | |
| "venue": "IEEE TRANSACTIONS ON SPEECH AND AUDIO PROCESSING", | |
| "volume": "8", | |
| "issue": "1", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Manhung Siu, Member, IEEE, and Mari Ostendorf, Senior Member, IEEE, \"Variable N-Grams and Extensions for Conversational Speech LanguageModeling\", IEEE TRANSACTIONS ON SPEECH AND AUDIO PROCESSING, VOL. 8, NO. 1, JANUARY 2000", | |
| "links": null | |
| }, | |
| "BIBREF9": { | |
| "ref_id": "b9", | |
| "title": "Interpolated Estimation of Markov Source Parameters From Sparse Data", | |
| "authors": [ | |
| { | |
| "first": "F", | |
| "middle": [], | |
| "last": "Jelinek", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "R", | |
| "middle": [ | |
| "L" | |
| ], | |
| "last": "Mercer", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1980, | |
| "venue": "Pattern Recognition in Practice", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "381--397", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "F. Jelinek and R.L. Mercer, \" Interpolated Estimation of Markov Source Parameters From Sparse Data,\" Pattern Recognition in Practice, E.S. Gelsema and L.N. Kanal, Eds., North-Holland Pub. Co., Amsterdam, pp. 381-397, 1980", | |
| "links": null | |
| }, | |
| "BIBREF11": { | |
| "ref_id": "b11", | |
| "title": "Headline generation based on statistical translation", | |
| "authors": [ | |
| { | |
| "first": "M", | |
| "middle": [], | |
| "last": "Banko", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "V", | |
| "middle": [], | |
| "last": "Mittal", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "M", | |
| "middle": [], | |
| "last": "Witbrock", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2000, | |
| "venue": "Proceedings of the 38th Annual Meeting of the Association for Computational Linguistics", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "318--325", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "M. Banko, V. Mittal and M. Witbrock, \"Headline generation based on statistical translation, \" in Proceedings of the 38th Annual Meeting of the Association for Computational Linguistics, 2000, pp. 318-325.", | |
| "links": null | |
| } | |
| }, | |
| "ref_entries": { | |
| "FIGREF1": { | |
| "text": "120 \u5c0f\u6642\uff0c\u6839\u64da\u6a19\u8a18\u6a94\u6848\uff0c\u53d6\u51fa\u4e3b\u64ad\u90e8\u5206\u56db\u5c0f\u6642\u4e09\u5341\u5206\u9418\uff0c \u5176\u4e2d\u4e09\u5c0f\u6642\u505a\u70ba\u8a13\uf996\u8a9e\uf9be\uff0c\u7d04 328MB\uff1b\u5269\u4e0b\u7d04\u4e00\u5c0f\u6642\u4e09\u5341\u5206\u9418\uff0c255 \u5247\u65b0\u805e\u5831\u5c0e\u4f5c\u70ba\u6e2c\u8a66\u8a9e\uf9be\uff0c\u7d04 166MB\u3002 \u5206\u5225\u8a08\u7b97\u97f3\u7bc0\u3001\u6bcd\u97f3\u548c\u5b57\u5143\u7684\u6b63\u78ba\uf961\uff0c\u6b63\u78ba\uf961\u7684\u8a08\u7b97\u6709\uff0c\u6b63\u78ba\uf961(accuracy)\u3001\u63d2\u5165\u932f\u8aa4(insertion)\u3001\u522a\u9664\u932f\u8aa4 (deletion)\u4ee5\u53ca\u66ff\u63db\u932f\u8aa4(substitution)\uff0c\u4e26\u4e14\u8003\u616e\u524dN \u540d\u8fa8\uf9fc\u7d50\u679c\u3002\u5176\u8a08\u7b97\u5f0f\u5982\u4e0b\uff1a --------------------------------------Syllable Results-------------------------------------------------------------------------Character Results-------------------------------------", | |
| "type_str": "figure", | |
| "uris": null, | |
| "num": null | |
| }, | |
| "TABREF0": { | |
| "html": null, | |
| "text": "\u8fa8\uf9fc\u6b63\u78ba\uf961\u6703\u5f71\u97ff\u6587\u7ae0\u8a9e\u610f\u7684\u5224\u65b7\uff0c\u70ba\uf9ba\u907f\u514d\u8aa4\u5224\u60c5\u5f62\u7684\u767c\u751f\uff0c\u7d93\u7531\u8a08\u7b97\u8fa8\uf9fc\u4fe1\u8cf4\u5206\uf969\uff0c\u53d6\u8fa8\uf9fc\u53ef\u4fe1\ufa01\u8f03 \u9ad8\u7684\u8a5e\u4f5c\u70ba\u6458\u8981\u3002\u7b2c\u4e09\u3001\u8a9e\u97f3\u6458\u8981\u8a5e\u8207\u8a5e\u4e4b\u9593\u7684\uf905\uf99a\u9593\u95dc\u4fc2\uff0c\u53ef\uf9dd\u7528\u8a9e\u8a00\u6a21\u578b\u5efa\uf9f7\u3002\u7b2c\u56db\u3001\u5206\u6790\u6587\u7ae0\u8a9e\u610f \u76f8\u4f9d\u7684\u95dc\u4fc2\uff0c\u5efa\u69cb\u5408\uf9e4\u7684\u8a9e\u610f\u4fee\u98fe\u95dc\uf997\u3002\u7b2c\u4e94\u3001\u914d\u5408\u6a5f\uf961\u5f0f\u6587\u6cd5\u898f\u5247\uff0c\u4f7f\uf906\u5b50\u5177\u6709\u6587\u6cd5\u898f\u5247\uff0c\uf9e0\u65bc\u95b1\uf95a\uf9e4 \u89e3\u3002\u6700\u5f8c\u4ee5\u52d5\u614b\u898f\u5283\u641c\u5c0b\u65b9\u6cd5\uff0c\u7522\u751f\u6700\u4f73\u7684\u6458\u8981\u8a5e\u7d44\u3002\u6b64\u5916\uff0c\u70ba\uf9ba\u4f7f\uf905\u63a5\u8a9e\u97f3\u5e73\uf904\u8f38\u51fa\uff0c\u5728\uf905\u63a5\u6458\u8981\u55ae\u5143 \u6642\uff0c\u5fc5\u9808\u8003\u616e\uf905\u63a5\uf9ca\u66a2\u548c\u5e73\uf904\u7684\u7a0b\ufa01\u3002\u56e0\u6b64\uff0c\u5728\u6458\u8981\u55ae\u5143\uf905\u63a5\u7684\u9078\u64c7\u4e0a\uff0c\u6211\u5011\u8003\u616e\u983b\u8b5c\u7279\u6027\uff1a\u5206\u5225\u4f7f\u7528\u983b \u8b5c\u4e2d\u5fc3(spectral centroid)\u3001\u983b\u8b5c\uf904\u52d5(spectral rolloff)\u3001\u983b\u8b5c\u8b8a\u9077(spectral flux)\u3001\u6642\u57df\u4e0a\u8d8a\uf9b2\uf961(time domain zero crossing, ZCR)\u548c\u6885\u723e\u5012\u983b\u8b5c\uf96b\uf969(Mel-frequency ceptral coefficient, MFCC) \uff0c\u627e\u51fa\u76f8\u9130\u5dee\uf962\u6700\u5c0f\u7684\uf905\u63a5\u55ae\u5143 \u4ee5\u751f\u6210\u5e73\uf904\u4e4b\u8a9e\u97f3\u8f38\u51fa\u3002 \u7136\u800c\uff0c\u5982\u4f55\u624d\u80fd\u5f9e\u8a9e\u97f3\u6587\u4ef6\u4e2d\u8403\u53d6\u51fa\u91cd\u8981\u7684\u8a5e\uf906\uff0c\u5efa\u69cb\u51fa\u80fd\u5920\u4ee3\u8868\u6587\u7ae0\u610f\u51fd\u7684\u5167\u5bb9\u3002\u672c\uf941\u6587\u5206\u5225\u5f9e\u8a9e \u97f3\u8072\u5b78(acoustics)\u3001\u8a9e\u8a00\u5b78(linguitic)\uff0c\uf906\u6cd5(syntax)\u548c\u8a9e\u610f(semantics)\u7b49\u65b9\u5411\u53bb\u89e3\u6c7a\u81ea\u52d5\u8a9e\u97f3\u6587\u4ef6\u6458\u8981\u53ef\u80fd\u9762", | |
| "num": null, | |
| "content": "<table><tr><td colspan=\"12\">\u5c0d\u7684\u554f\u984c\uff0c\u4e00\u7bc7\u8a9e\u97f3\u6587\u4ef6\u900f\u904e\u7279\u5fb5\uf96b\uf969\u7684\u8a08\u7b97\uff0c\u53ef\u88ab\u5206\u6790\u6210\u4e94\u500b\u4e3b\u8981\u7684\u7279\u5fb5\u5206\uf969\uff0c\u5305\u542b\u6709\uff1a(1) \u8a9e\u97f3\u8fa8\uf9fc</td></tr><tr><td colspan=\"7\">\u4fe1\u8cf4 (confidence measure,</td><td colspan=\"3\">( m C w )</td><td colspan=\"2\">) \u5206\uf969\uff1b(2)\u5b57\u8a5e\u76f8\u5c0d\u65bc\u6587\u7ae0\u6240\u4ee3\u8868\u7684\u91cd\u8981\u6027 (word significance,</td><td>( ) m R w</td><td>) \u5206</td></tr><tr><td colspan=\"12\">\uf969\uff1b(3)\u8a9e\u8a00\u5b78\u7d50\u69cb\u76f8\u9130 (linguistic trigram,</td><td>( | m L w w m</td><td>\u2212</td><td>2</td><td>,</td><td>m w</td><td>) \u2212 ) \u5206\uf969\uff1b(4)\u8a9e\u610f\u76f8\u4f9d\u6cd5\u5247 (semantic dependency 1</td></tr><tr><td>grammars,</td><td>SDG B</td><td>(</td><td>1 w w , m m \u2212</td><td>)</td><td colspan=\"7\">) \u5206\uf969\uff1b\u4ee5\u53ca(5)\u6a5f\uf961\u5f0f\u6587\u6cd5\u898f\u5247 (probabilistic context-free grammars,</td><td>( ) P S ) \u5206\uf969\u3002</td></tr><tr><td colspan=\"12\">\u56e0\u70ba\u5206\uf969\u503c\u57df\u5927\u5c0f\u4e26\uf967\u76f8\u540c\uff0c\u6240\u4ee5\u6211\u5011\u5206\u5225\u8a08\u7b97\u5206\uf969\u7684\u6700\u5927\u503c</td><td>score Max \u548c\u6700\u5c0f\u503c</td><td>score Min \uff0c\u4f9d\u5176\uf967\u540c\u503c\u57df\u5c0d</td></tr><tr><td colspan=\"4\">\u5404\u5206\uf969 score X \u505a\u6b63\u898f\u5316 (</td><td colspan=\"2\">X</td><td colspan=\"2\">score</td><td>\u2212</td><td colspan=\"2\">score Min</td><td>) / (</td><td>score Max</td><td>\u2212</td><td>score Min</td><td>)</td><td>\u5c07\u6bcf\u4e00\u500b\u5206\uf969\u503c\u8abf\u6574\u70ba\u5f9e 0 \u5230 1 \u4e4b\u9593\u3002\u8a9e\u97f3\u6587\u4ef6</td></tr><tr><td colspan=\"12\">\u7d93\u904e\u5927\u8a5e\u5f59\uf99a\u7e8c\u8a9e\u97f3\u8fa8\uf9fc\uff0c\u7522\u751f\u4e00\u7bc7\u8a5e\u9577\u70ba M \u7684\u8f49\u8b6f\u6587\u4ef6</td><td>1 { , , ,..., 2 3</td><td>M</td><td>1</td><td>, } M</td></tr></table>", | |
| "type_str": "table" | |
| }, | |
| "TABREF1": { | |
| "html": null, | |
| "text": "\u8868\u793a frequency count\u3002\u4f46\u70ba\uf9ba\u907f\u514d\u8a31\u591a\u8a5e\u7d71\u8a08\uf967\u5230 trigram \u60c5\u6cc1\u767c\u751f\uff0c\u5f15\u7528 Jelinek et al.\u6240\u63d0\u51fa\u7684\u5e73\uf904\u65b9\u6cd5(N gram smoothing)[10]\uff0c\u5167\u63d2 trigram, bigram \u548c unigram \u7b49 \u76f8\u95dc\u6a5f\uf961\u503c\u3002\u8868\u793a\u65b9\u6cd5\u5982\u4e0b\uff1a", | |
| "num": null, | |
| "content": "<table><tr><td/><td/><td/><td/><td/><td/><td/><td/><td>)</td></tr><tr><td>2.3 \u8a9e\u8a00\u5b78\u7d50\u69cb\u76f8\u9130\u5206\uf969</td><td/><td/><td/><td/><td/><td/><td/></tr><tr><td>\u6211\u5011\uf9dd\u7528\u4e09\uf99a\u8a9e\u8a00\u6a21\u578b</td><td>3 ( | , ) 1 2 L w w w</td><td>=</td><td colspan=\"5\">1 ( , , )/ ( , ) 2 3 1 2 F w w w F w w</td><td>\uff0c\u5efa\uf9f7\u8a5e\u8207\u8a5e\u4e4b\u9593\u76f8\u63a5\u7684\u60c5\u6cc1\uff0c\u4f7f</td></tr><tr><td colspan=\"4\">\u6458\u8981\u6700\u5f8c\u7d50\u679c\uf901\u52a0\u7b26\u5408\u8a9e\u8a00\u5b78\u7d50\u69cb[9]\u3002\u5176\u4e2d ( ) F i 3 1 2 1 1 2 3 1</td><td>2</td><td>2</td><td>1</td><td>2</td><td>1</td><td>3</td><td>1</td></tr></table>", | |
| "type_str": "table" | |
| }, | |
| "TABREF6": { | |
| "html": null, | |
| "text": "\u6458\u8981\u4e4b\u5404\u5206\uf969\u91cd\u8981\u6027\u8a55\u4f30 \u7531\u5be6\u9a57\u7d50\u679c(\u5716 8)\u53ef\u77e5\uff0c\uf9dd\u7528\u6c42\u53d6\u95dc\u9375\u8a5e\u7684\u4f5c\u6cd5(word significance score)\u6548\u679c\u6700\u70ba\u986f\u8457\uff0c\u5176\u6b21\u70ba\u8a9e\u610f\u76f8\u4f9d\u6cd5 \u5247\u3001\u4e09\uf99a\u8a9e\u8a00\u6a21\u578b\uff0c\u6700\u5f8c\u662f\u8a9e\u97f3\u8fa8\uf9fc\u4fe1\u8cf4\u5206\uf969\u3002 ALL \u4ee3\u8868\u7d50\u5408\u56db\u7a2e\u5206\uf969\u6240\u5f97\u5230\u7684\u6458\u8981\u7d50\u679c\uff0c\u4f9d\u64da\u5404\u7a2e\u77e5\uf9fc\u6240\u4ee3\u8868\u7684\u91cd\u8981\u6027\u7a0b\ufa01\uff0c\u8a2d\u5b9a\u5176\u6b0a\u91cd\u5206\u5225\u70ba C(0.1)\u3001L(0.2)\u3001W(0.4)\u548c S(0.3)\uff0c\u8a55\u4f30\u6b63\u78ba\uf961 accuracy \u70ba\u767e\u5206\u4e4b\u4e09\u5341\u4e94\u3002\u8a73\u7d30\u7684\u5be6\u9a57\u7d50\u679c\u5982(\u5716 9)\u6240\u793a\u3002\u7531 (\u5716 9)\u5f97\u77e5\uff0c\u6458\u8981\u932f\u8aa4\u8f03\u5e38\u767c\u751f\u5728\u63d2\u5165\u932f\u8aa4\uff0c\u5176\u6b21\u70ba\u66ff\u63db\u932f\u8aa4\u548c\u522a\u9664\u932f\u8aa4\u3002\u7531\u6b64\u53ef\u77e5\u6458\u8981\u7d50\u679c\u7684\u597d\u58de\uff0c\u4e3b\u89c0 \u56e0\u7d20\u5f71\u97ff\u8f03\u5927\uff0c\u63d2\u5165\u548c\u66ff\u63db\u932f\u8aa4\u8f03\u5bb9\uf9e0\u767c\u751f\u3002", | |
| "num": null, | |
| "content": "<table><tr><td>) \uff1a\u5c0d\u65bc \u7684\u67e5\u8a62\u7d50\u679c\uff0c \u6392\u5e8f\u7b2c \u7bc7\u76f8\u95dc\u6587\u7ae0\u3002mAP \u53ef\u4ee5\u5206\u6790\u67e5\u8a62\u7d50\u679c\uff0c\u662f\u5426\u6709\u6b63\u76f8\u95dc\u6027\uff0c\u4e5f\u5c31\u662f\u524d\u9762\u7684\u6587\u7ae0\u662f\u76f8\u95dc\u7684\uff0c\u800c\u5f8c\u9762 \u5176\u4e2d\uff0c \uff1a\u67e5\u8a62\u7684\u554f\uf906\uf969\u3002 \uff1a\u5c0d\u65bc \u7684\u67e5\u8a62\u7d50\u679c\uff0c\u5171\u6709\u5e7e\u7bc7\u76f8\u95dc\u6587\u7ae0\u3002 \u7684\u6587\u7ae0\u53ef\u80fd\u76f8\u95dc\u6027\u8f03\u4f4e\uff0cmAP \u66f2\u7dda\uf974\u7121\u8df3\u52d5\u7684\u60c5\u5f62\uff0c\u5247\u8868\u793a\u8a55\u4f30\u67e5\u8a62\u7684\u6548\u679c\u597d\uff0c\u53cd\u4e4b\u4ea6\u7136\u3002rAP \u5247\u53ef\u4ee5 \u5224\u65b7\u5728\u7b2c\u5e7e\u7bc7\u6587\u4ef6\uff0c\u6587\u7ae0\u5c0d\u65bc\u67e5\u8a62\u7d50\u679c\u76f8\u95dc\ufa01\u7684\ufa09\u4f4e\u3002\u7531(\u5716 7)\u89c0\u5bdf\u5f97\u77e5\uff0c\u7576\u6458\u8981\u6bd4\uf9b5\u8d8a\u9ad8\u5247\u6240\u542b\u7684\u8cc7\u8a0a\u8d8a \u9ad8\uff0c\u4e5f\u5c31\u662f\u8cc7\u8a0a\u58d3\u7e2e\u8d8a\u5c0f\u5247\u8a9e\u610f\u4fdd\uf9cd\u7a0b\ufa01\u8d8a\u9ad8\u3002\u4f46\u662f\uff0c\u7576\u6211\u5011\u505a 30%\u7684\u6458\u8981\u6642\uff0c\u6240\u6aa2\uf96a\u7684\u524d\u56db\u7bc7\u6587\u4ef6\u8207\u6458 \u8981 70%\u548c 50%\u6642\u7684\u7d50\u679c\u5f88\u76f8\u8fd1\u3002 q N i N i q ik rank i q k \u53e6\u5916\uff0c\u5c07\u6e2c\u8a66\u97f3\u6a94\u505a\u4eba\u5de5\u7684\u6458\u8981\u8a18\uf93f\u5f8c\uff0c\u8207\u81ea\u52d5\u6458\u8981\u7d50\u679c\u76f8\u5c0d\u7167\uff0c\u5206\u5225\u8a08\u7b97\u5176\u6b63\u78ba\uf961\u3001\u63d2\u5165\u932f\u8aa4\u3001\u522a \u9664\u932f\u8aa4\u4ee5\u53ca\u66ff\u63db\u932f\u8aa4\u7b49\uff0c\u5982(\u5f0f 17)\u3002\u540c\u6642\uff0c\u5be6 0 \u9a57\u5404\u7a2e\u77e5\uf9fc\u5eab\u6240\u4ee3\u8868\u7684\u91cd\u8981\u7a0b\ufa01\uff0c\u4ee5(C_L_W_S)\u5206\u5225\u4ee3\u8868 \u8a9e\u97f3\u8fa8\uf9fc\u4fe1\u8cf4\u5206\uf969\u3001\u8a9e\u8a00\u5b78\u5206\uf969\u3001\u8a5e\u91cd\u8981\u6027\u5206\uf969\u548c\u8a9e\u610f\u76f8\u4f9d\u6cd5\u5247\u5206\uf969\uff0c\u8003\u616e\u5404\u7a2e\u60c5\u6cc1\u5982\u4e0b\u5716\u6240\u793a\uff1a 0.4 0.5 0.6 0.7 0.8 0.9 1 1 2 3 4 5 6 7 8 9 1 0 Num b e r o f Do c um e nts R e trie ve d Ave ra g e P re c is io n ORG m AP ORG rAP S UM30% m AP S UM30% rAP S UM50% m AP S UM50% rAP S UM70% m AP S UM70% rAP \u5716 7. \u91cd\u8981\u8cc7\u8a0a\u6aa2\uf96a\u7684\u7d50\u679c \uf905\u63a5\u8a9e\u97f3\u7684\u5be6\u9a57\u53ef\u7531(\u5716 10)\u8868\u793a\uff0c\u8acb\u5341\u4f4d\u53d7\u6e2c\u8005\u5206\u5225\u91dd\u5c0d\uf967\u540c\u6458\u8981\u6bd4\uf9b5\u8a55\u6bd4\u3002\u53d7\u6e2c\u8005\u5148\u770b\u904e\u539f\u59cb\u6a19\u6e96 \u5831\u5c0e\uff0c\u4e26\uf9b0\u807d\u5831\u5c0e\u5167\u5bb9\u4e4b\u5f8c\u3002\u6bd4\u8f03\u6458\u8981\u5f8c\u7684\u6587\u5b57\u7d50\u679c\u548c\uf9b0\u807d\u8a9e\u97f3\uf905\u7d50\u6548\u679c\uff0c\u662f\u5426\u80fd\u8868\u9054\u5831\u5c0e\u6587\u610f\u53ca\u5408\u6210\u8a9e \u97f3\u662f\u5426\uf9ca\u66a2\uff0c\u8a55\u6bd4\u4e00\u5230\u5341\u5206\uf969\uff0c\u4ee3\u8868\u5f9e\uf99d\u5230\u512a\u7684\u8868\u73fe\u6548\u679c\u3002 \u5716 9. \u6458\u8981\u7d50\u679c\u6b63\u78ba\uf961\u8a55\u4f30 \u5716 10. \u6458\u8981\uf905\u63a5\u53ca\u5408\u6210\u7d50\u679c\u8a55\u4f30 5. \u7d50\uf941\u53ca\u672a\uf92d\u5c55\u671b \u672c\uf941\u6587\u63d0\u51fa\u65b0\u805e\u8a9e\uf9be\u5eab\u53ca\u8a9e\u610f\u76f8\u4f9d\u6cd5\u5247\u65bc\u4e2d\u6587\u8a9e\u97f3\u6587\u4ef6\u6458\u8981\uff0c\u540c\u6642\u5c0d\u8a9e\u97f3\uf905\u63a5\u55ae\u5143\u8a08\u7b97\u983b\u8b5c\u4e0a\u7684\u7279\u5fb5 \uf96b\uf969\uff0c\uf9dd\u7528\u52d5\u614b\u898f\u5283\u641c\u5c0b\u65b9\u6cd5\uff0c\u751f\u6210\u4e00\u500b\u517c\u5177\u8a9e\u610f\u58d3\u7e2e\u548c\u807d\u89ba\u6548\u679c\uf9ca\u66a2\u7684\u6458\u8981\u7d50\u679c\u3002\u5206\u6790\u6458\u8981\u8a9e\u97f3\u6587\u4ef6\u7684 \u8072\u5b78\u3001\u8a9e\u610f\u548c\u8a9e\u6cd5\u7b49\u7279\u5fb5\uff0c\u7d50\u5408\u8a9e\u97f3\u8fa8\uf9fc\u4fe1\u8cf4\u5206\uf969\u3001\u8a5e\u91cd\u8981\u6027\u5206\uf969\u3001\u8a9e\u8a00\u5b78\u5206\uf969\u3001\uf906\u6cd5\u7d50\u69cb\u5206\uf969\u53ca\u8a9e\u610f\u76f8 \u4f9d\u6cd5\u5247\u5206\uf969\u3002\u6458\u8981\u55ae\u5143\uf905\u7d50\u5f9e\u983b\u8b5c\u4e0a\u53d6\u4e94\u9805\u7279\u5fb5\uf96b\uf969\uff0c\u983b\u8b5c\u4e2d\u5fc3\u3001\u983b\u8b5c\uf904\u52d5\u3001\u983b\u8b5c\u8b8a\u9077\u3001\u8d8a\uf9b2\uf961\u4ee5\u53ca\u6885\u723e \u5012\u983b\u8b5c\uf96b\uf969\uff0c\u6c7a\u5b9a\u6700\u4f73\u7684\u8a9e\u97f3\uf905\u63a5\u3002\u76ee\u524d\u5728\u516b\u6210\u7684\u8a9e\u97f3\u8fa8\uf9fc\uf961\u4e0b\uff0c\u5be6\u9a57\u8b49\u5be6\u7cfb\u7d71\u53ef\u4ee5\u505a\u5230\uf97c\u597d\u7684\u8a9e\u610f\u64f7\u53d6 \u4fdd\uf9cd\uff0c\u4ee5\u53ca\uf9ca\u66a2\u7684\u6458\u8981\u8a9e\u97f3\u6548\u679c\u3002 \u8a9e\u97f3\u6458\u8981\u7684\u76ee\u7684\uff0c\u65e8\u5728\u58d3\u7e2e\u8a9e\u97f3\u6587\u4ef6\uff0c\u53d6\u51fa\u5177\u4ee3\u8868\u6027\u5167\u5bb9\uff0c\u4e26\u4e14\u80fd\uf9ca\u66a2\u5730\u5c07\u8a9e\u97f3\uf905\u63a5\u8f38\u51fa\u3002\u4ee5\u6b64\u7814\u7a76 \u70ba\u57fa\u790e\u5c55\u671b\u672a\uf92d\uff0c\u53ef\u85c9\u7531\uf997\u5408\u5404\u7a2e\u65b9\u6cd5\uff0c\u63a2\u8a0e\u5982\u4f55\u6539\u5584\u6458\u8981\u6548\u679c\uff1a 1) \u5f9e\u6458\u8981\u8a9e\u97f3\u53ef\u5206\u70ba\u6587\u9ad4\u898f\u7bc4\u5f0f\u8a9e\u97f3\u548c\u81ea\u7136\u53e3\u8a9e\u5f0f\u8a9e\u97f3\uf978\u5927\uf9d0\u3002\u5176\u4e2d\uff0c\u6587\u9ad4\u898f\u7bc4\u5f0f\u8a9e\u97f3\u662f\u6307\u8a9e\u97f3 \u5167\u5bb9\u6709\u4e8b\u5148\u7d93\u904e\u8a2d\u8a08\uff0c\u8868\u9054\u5167\u5bb9\u8207\u66f8\u672c\u6216\u6587\u7ae0\u7684\u683c\u662f\u76f8\u8fd1\uff0c\u50cf\u662f\u65b0\u805e\u5831\u5c0e\u3002\u800c\u81ea\u7136\u53e3\u8a9e\u5f0f\u8a9e\u97f3 \u5247\u6307\u8a9e\u97f3\u5167\u5bb9\u7121\u7d93\u904e\u8a2d\u8a08\uff0c\u8868\u9054\u5167\u5bb9\u662f\uf9f6\u6642\u601d\u8003\u61c9\u5c0d\uff0c\u50cf\u662f\u5c0d\u8a71\u3001\u8a2a\u8ac7\u7b49\u3002 2) \u5206\u6790\u6587\u7ae0\u8a9e\u610f\uff0c\u9032\u3127\u6b65\u63a2\u8a0e\u61c9\u7528 Ontology \u65bc\u6458\u8981\u3002 3) \u4ee5\u65b0\u805e\u8a9e\u97f3\u70ba\uf9b5\uff0c\u53ef\u5c07\u65b0\u805e\u5206\uf9d0\u4e26\u4f9d\u7167\uf967\u540c\u7684\u65b0\u805e\uf9d0\u5225\uff0c\u62bd\u53d6\u51fa\u5177\u4ee3\u8868\u6027\u7684\u95dc\u9375\u8a5e\uff0c\u6216\u5efa\uf9f7\uf967 \u540c\u65b0\u805e\uf9d0\u5225\u7684\uf906\u6cd5\u7d50\u69cb\u6a21\u7d44\uff0c\u4ee5\u8f14\u52a9\u6458\u8981\u751f\u6210\u3002 4) \u5206\u6790\u8a9e\u97f3\u8072\u5b78\u4e0a\u7279\u6027\uff0c\u5982\uff1a\u97f3\u9ad8\u3001\u9031\u671f\u548c\u80fd\uf97e\u7b49\u3002 5) \u85c9\u7531\u7db2\u969b\u7db2\uf937\u7684\u5e6b\u52a9\uff0c\u53ef\u5206\u6790\u56e0\u70ba\u6642\u9593\u7684\u63a8\u9032\uff0c\u6240\u7522\u751f\u7684\u65b0\u8a5e\u3001\u6587\u7ae0\u7528\u6cd5\u7684\u8868\u9054\uff0c\u548c\u5404\uf9b4\u57df\u7684 \u77e5\uf9fc\u7b49\u3002 \u8a8c\u8b1d \u5716 8. 4.3 \uf905\u63a5\u6548\u679c\u8a55\u4f30 \u611f\u8b1d\u570b\u79d1\u6703\u652f\u6301\u672c\u7814\u7a76\u8a08\u756b\uff0c\u8a08\u756b\u7de8\u865f NSC90-2213-E-006-088\u3002</td></tr></table>", | |
| "type_str": "table" | |
| } | |
| } | |
| } | |
| } |