{ "paper_id": "O13-1003", "header": { "generated_with": "S2ORC 1.0.0", "date_generated": "2023-01-19T08:03:44.423633Z" }, "title": "Using Speech Assessment Technique for the Validation of Taiwanese Speech Corpus", "authors": [ { "first": "Yu -Jhe", "middle": [], "last": "\u674e\u6bd3\u54f2", "suffix": "", "affiliation": {}, "email": "" }, { "first": "", "middle": [], "last": "Li", "suffix": "", "affiliation": {}, "email": "" }, { "first": "Chung-Che", "middle": [], "last": "\u738b\u5d07\u5586", "suffix": "", "affiliation": {}, "email": "" }, { "first": "", "middle": [], "last": "Wang", "suffix": "", "affiliation": {}, "email": "" }, { "first": "Liang-Yu", "middle": [], "last": "Chen", "suffix": "", "affiliation": {}, "email": "" }, { "first": "Jyh-Shing", "middle": [ "Roger" ], "last": "Jang", "suffix": "", "affiliation": {}, "email": "jang@mirlab.org" }, { "first": "Ren-Yuan", "middle": [], "last": "Lyu", "suffix": "", "affiliation": {}, "email": "renyuan.lyu@gmail.com" }, { "first": "", "middle": [], "last": "\u9577\u5e9a\u5927\u5b78\u8cc7\u8a0a\u5de5\u7a0b\u5b78\u7cfb", "suffix": "", "affiliation": {}, "email": "" } ], "year": "", "venue": null, "identifiers": {}, "abstract": "", "pdf_parse": { "paper_id": "O13-1003", "_pdf_hash": "", "abstract": [], "body_text": [], "back_matter": [], "bib_entries": {}, "ref_entries": { "TABREF0": { "type_str": "table", "content": "
Proceedings of the Twenty-Fifth Conference on Computational Linguistics and Speech Processing (ROCLING 2013)
\u5225\u70ba\uff1a40.22%\u300141.21%\u300144.35%\u3002
\u7531\u7d50\u679c\u770b\u4f86\uff0c\u7d93\u904e\u7be9\u9078\u5f8c\u8a9e\u6599\u6240\u8a13\u7df4\u51fa\u7684\u8072\u5b78\u6a21\u578b\u8207\u672a\u7d93\u7be9\u9078\u8a9e\u6599\u6240\u7522\u751f\u7684\u8072\u5b78\u6a21\u578b\uff0c
\u5176\u8fa8\u8b58\u7387\u7684\u5dee\u5225\u6700\u9ad8\u53ef\u9054 4.13%\uff0c\u8b49\u5be6\u672c\u8ad6\u6587\u6240\u63d0\u7684\u65b9\u6cd5\uff0c\u85c9\u7531\u8a9e\u97f3\u8a55\u5206\u78ba\u5be6\u80fd\u6709\u6548\u7684
\u81ea\u52d5\u7be9\u9078\u6389\u6709\u554f\u984c\u7684\u8a9e\u53e5\u3002
\u8072\u5b78\u6a21\u578b \u8a13\u7df4\u968e\u6bb5\u8a13\u7df4\u8a9e\u6599\u7279\u5fb5\u64f7\u53d6\u6a21\u578b\u8a13\u7df4\u8072\u5b78\u6a21\u578b
\u8a9e\u97f3\u8a55\u5206\u9ad8\u5206\u5340
\u968e\u6bb5\u5f85\u6574\u7406\u7279\u5fb5\u64f7\u53d6\u8a9e\u97f3\u8a55\u5206\u4e2d\u9593\u503c\u5340
\u8a9e\u6599\u4f4e\u5206\u5340
\u4e8c\u6b21\u6aa2\u9a57 \u968e\u6bb5\u4f4e\u5206\u5340\u4eba\u5de5\u6a19\u8a18 \u6458\u8981\u53ef\u7528\u8a9e\u6599
\u672c\u8ad6\u6587\u7684\u4e3b\u8981\u7814\u7a76\u70ba\u4f7f\u7528\u8a9e\u97f3\u8fa8\u8b58\u53ca\u7d50\u5408\u8a9e\u97f3\u8a55\u5206\uff0c\u5c0d\u672a\u6574\u7406\u7684\u53f0\u8a9e\u8a9e\u6599\u9032\u884c\u521d\u6b65\u7684\u7be9 \u7279\u5fb5\u64f7\u53d6 SVM \u5206\u985e \u8a9e\u6599
\u9078\u3002\u85c9\u7531\u6a5f\u5668\u5148\u904e\u6ffe\u6389\u6709\u554f\u984c\u7684\u97f3\u6a94\uff0c\u5982\u9304\u97f3\u97f3\u91cf\u904e\u5c0f\u3001\u592a\u591a\u96dc\u8a0a\u3001\u9304\u97f3\u97f3\u6a94\u5167\u5bb9\u6709\u8aa4 \u5668 \u4e0d\u826f\u8a9e\u6599
\u7b49\u60c5\u5f62\uff0c\u53d6\u4ee3\u50b3\u7d71\u4eba\u5de5\u807d\u6e2c\u8cbb\u6642\u7684\u505a\u6cd5\u3002\u672c\u8ad6\u6587\u5982\u5716\u4e00\u6240\u793a\uff0c\u53ef\u5206\u70ba\u4e09\u500b\u968e\u6bb5\uff0c\u5206\u5225\u662f\uff1a
\u300c\u57fa\u790e\u8072\u5b78\u6a21\u578b\u8a13\u7df4\u300d \u3001 \u300c\u8a9e\u97f3\u8a55\u5206\u8207\u932f\u8aa4\u539f\u56e0\u6a19\u8a18\u300d\u53ca\u300c\u6548\u80fd\u8a55\u4f30\u300d \u3002 \u5716\u4e00\u3001\u8a9e\u6599\u6574\u7406\u7cfb\u7d71\u6d41\u7a0b\u5716
\u65bc\u57fa\u790e\u8072\u5b78\u6a21\u578b\u8a13\u7df4\u968e\u6bb5\uff0c\u4ee5\u9577\u5e9a\u5927\u5b78\u63d0\u4f9b\u7684\u53f0\u8a9e\u8a9e\u6599 ForSD (Formosa Speech Database) [1] \u70ba\u6750\u6599\uff0c\u4f7f\u7528\u96b1\u85cf\u5f0f\u99ac\u53ef\u592b\u6a21\u578b(Hidden Markov Model, HMM)\u3001 \u6885\u723e\u5012 \u95dc\u9375\u8a5e\uff1a\u53f0\u8a9e\u8a9e\u6599\u6574\u7406\u3001\u96b1\u85cf\u5f0f\u99ac\u53ef\u592b\u6a21\u578b\u3001\u8a9e\u97f3\u8a55\u5206\u3001\u8a9e\u97f3\u8fa8\u8b58\u3001\u652f\u6301\u5411\u91cf\u6a5f
Keywords: Taiwanese Corpus Validation, Hidden Markov model, Speech Assessment, \u983b\u8b5c\u4fc2\u6578(Mel-frequency Cepstral Coefficients, MFCCs) [2] \u548c\u5c0d\u6578\u80fd\u91cf(Log energy) \u505a\u70ba\u8a9e\u97f3\u7279\u5fb5\u9032\u884c\u8072\u5b78\u6a21\u578b\u7684\u8a13\u7df4\u3002\u8072\u5b78\u6a21\u578b\u55ae\u4f4d\u5206\u5225\u70ba\uff1a\u55ae\u97f3\u7d20\u8072\u5b78\u6a21\u578b (Monophone Support Vector Machine.
acoustic model) \u3001\u97f3\u7bc0\u5167\u53f3\u76f8\u95dc\u96d9\u9023\u97f3\u7d20\u8072\u5b78\u6a21\u578b(Biphone acoustic model)\u53ca\u97f3\u7bc0\u5167 \u5de6\u53f3\u76f8\u95dc\u4e09\u9023\u97f3\u7d20\u8072\u5b78\u6a21\u578b(Triphone acoustic model) \uff0c\u5176\u91dd\u5c0d\u6e2c\u8a66\u8a9e\u6599\u9032\u884c\u81ea\u7531\u97f3\u7bc0 \u53c3\u8003\u6587\u737b
\u89e3\u78bc\u8fa8\u8b58\u7db2\u8def(Free syllable decoding)\u7684\u97f3\u7bc0\u8fa8\u8b58\u7387(Syllable accuracy)\u6700\u4f73\u7d50\u679c\u5206 [1] Ren-yuan Lyu, Min-siong Liang, Yuang-chin Chiang, Toward Construction A \u5225\u70ba\uff1a27.20%\u300143.28%\u300145.93%\u3002\u5176\u4e2d\u5de6\u53f3\u76f8\u95dc\u4e09\u9023\u97f3\u7d20\u8072\u5b78\u6a21\u578b\u7684\u8fa8\u8b58\u7387\u6700\u4f73\uff0c\u56e0 Multilingual Speech Corpus for Taiwanese (Min-nan), Hakka, and Mandarin, \u6b64\u6211\u5011\u9078\u64c7\u6b64\u6a21\u578b\u9032\u884c\u7b2c\u4e8c\u968e\u6bb5\u7684\u5be6\u9a57\u3002 International Journal of Computational Linguistics and Chinese Language Processing,
\u65bc\u8a9e\u97f3\u8a55\u5206\u8207\u932f\u8aa4\u539f\u56e0\u6a19\u8a18\u968e\u6bb5\uff0c\u5c07\u65bc\u57fa\u790e\u8072\u5b78\u6a21\u578b\u8a13\u7df4\u968e\u6bb5\u5df2\u8a13\u7df4\u597d\u7684\u5de6\u53f3\u76f8\u95dc\u4e09 2004.
\u9023\u97f3\u7d20\u8072\u5b78\u6a21\u578b\uff0c\u5c0d\u5f85\u6574\u7406\u7684\u8a9e\u6599\u9032\u884c\u8a9e\u97f3\u8a55\u5206 [3, 4]\u3002\u8a9e\u97f3\u8a55\u5206\u80fd\u85c9\u7531\u8072\u5b78\u6a21\u578b\u5c0d\u9304 [2] Steven B. Davis and Paul Mermelstein, Comparison of Parametric Representation for \u97f3\u9032\u884c\u8a55\u5206\uff0c\u5728\u672c\u8ad6\u6587\u4e2d\u4ee5\u8a55\u5206\u5f8c\u7684\u5206\u6578\u4f86\u8a55\u91cf\u97f3\u6a94\u7684\u8207\u6587\u672c\u9593\u7684\u76f8\u4f3c\u7a0b\u5ea6\u3002\u4f46\u4f9d\u64da\u524d Monosyllabic Word Recognition in Continuously Spoken Sentences, IEEE International
\u4eba\u7814\u7a76 [5]\uff0c\u5728\u67d0\u4e9b\u72c0\u6cc1\u4e0b\u8a9e\u97f3\u8a55\u5206\u7684\u5206\u6578\u4e26\u4e0d\u5408\u7406\uff0c\u56e0\u6b64\u5728\u672c\u8ad6\u6587\u4e2d\uff0c\u70ba\u4e86\u964d\u4f4e\u8a55\u5206 Conference on Acoustics, 1980.
\u6642\u4e0d\u5408\u7406\u60c5\u5f62\u51fa\u73fe\u7684\u6a5f\u7387\uff0c\u52a0\u5165\u4e86\u4e09\u7a2e\u5206\u6578\u8abf\u6574\u7684\u6263\u5206\u6a5f\u5236\uff0c\u5206\u5225\u662f\uff1a\u97f3\u7bc0\u4e4b\u97f3\u6846\u500b\u6578 \u5dee\u8ddd\u904e\u5927\u3001\u97f3\u7bc0\u4e2d\u9023\u7e8c\u97f3\u7d20\u4e4b\u97f3\u6846\u6578\u76ee\u904e\u5c0f\u3001\u4ee5\u53ca\u6587\u672c\u8207\u8fa8\u8b58\u7d50\u679c\u4e4b\u97f3\u7bc0\u6578\u76ee\u4e0d\u4e00\u3002\u800c [3] \u674e\u4fca\u6bc5\uff0c\u8a9e\u97f3\u8a55\u5206\uff0c\u6e05\u83ef\u5927\u5b78\u78a9\u58eb\u8ad6\u6587\uff0c\u6c11\u570b 91 \u5e74\u3002
\u6b64\u8a55\u5206\u7d50\u679c\u5c07\u4f9d\u7167\u9580\u6abb\u503c\u5206\u70ba\u4e09\u90e8\u5206\uff0c\u5206\u5225\u70ba\u4f4e\u5206\u5340\u3001\u4e2d\u9593\u503c\u5340\u53ca\u9ad8\u5206\u5340\u3002\u4e14\u91dd\u5c0d\u4f4e\u5206 [4] \u9673\u5b8f\u745e\uff0c\u4f7f\u7528\u591a\u91cd\u8072\u5b78\u6a21\u578b\u4ee5\u6539\u826f\u53f0\u8a9e\u8a9e\u97f3\u8a55\u5206\uff0c\u6e05\u83ef\u5927\u5b78\u78a9\u58eb\u8ad6\u6587\uff0c\u6c11\u570b 100 \u5e74\u3002
\u5340\u90e8\u5206\u8a9e\u6599\u9032\u884c\u4eba\u5de5\u6a19\u8a18\uff0c\u6a19\u8a18\u5176\u932f\u8aa4\u539f\u56e0\uff0c\u518d\u5c0d\u5176\u64f7\u53d6\u7279\u5fb5\uff0c\u4f7f\u7528\u652f\u6301\u5411\u91cf\u6a5f (Support [5] \u9ec3\u6b66\u986f\uff0c\u57fa\u65bc 32 \u4f4d\u5143\u6574\u6578\u904b\u7b97\u8655\u7406\u5668\u4e4b\u83ef\u8a9e\u8a9e\u97f3\u8a55\u5206\u7684\u6539\u826f\u8207\u7814\u7a76\uff0c\u6e05\u83ef\u5927\u5b78\u78a9 Vector Machine, SVM)\u8a13\u7df4\u51fa\u5206\u985e\u5668\uff0c\u6700\u5f8c\u4ee5\u8a72\u5206\u985e\u5668\u5c0d\u4f4e\u5206\u5340\u8a9e\u6599\u9032\u884c\u4e8c\u6b21\u6aa2\u9a57\uff0c\u5c07 \u4f4e\u5206\u5340\u8a9e\u6599\u5206\u70ba\u53ef\u7528\u8a9e\u6599\u53ca\u4e0d\u826f\u8a9e\u6599\u3002 \u58eb\u8ad6\u6587\uff0c\u6c11\u570b 96 \u5e74\u3002
\u65bc\u6548\u80fd\u8a55\u4f30\u968e\u6bb5\uff0c\u5c07\u539f\u5148\u8a13\u7df4\u8a9e\u6599\u5206\u5225\u52a0\u5165 \u300c\u672a\u6574\u7406\u8a9e\u6599\u300d \u3001 \u300c\u4e2d\u9593\u503c\u5340\u53ca\u9ad8\u5206\u5340\u8a9e\u6599\u300d \u3001
\u300c\u9ad8\u5206\u5340\u8a9e\u6599\u300d\u9032\u884c\u8072\u5b78\u6a21\u578b\u7684\u8a13\u7df4\uff0c\u6bd4\u8f03\u7be9\u9078\u8a9e\u6599\u524d\u3001\u5f8c\u6548\u80fd\uff0c\u5176\u97f3\u7bc0\u8fa8\u8b58\u7387\u7d50\u679c\u5206
", "text": "Proceedings of the Twenty-Fifth Conference on Computational Linguistics and Speech Processing(ROCLING 2013)", "html": null, "num": null } } } }