Non-native English speaker's speech correction, based on domain focused document

カテゴリ: 部門大会

論文No: SS4-7

グループ名: 【C】平成28年電気学会電子・情報・システム部門大会講演論文集

発行日: 2016/08/31

タイトル(英語): Non-native English speaker's speech correction, based on domain focused document

著者名: Radzikowski Kacper(早稲田大学),Wang Le(早稲田大学),吉江 修(早稲田大学)

著者名(英語): Kacper Radzikowski|Le Wang|Osamu Yoshie

キーワード: BM25F|HMM|音声認識修正|文訂正|キーワード専門文書|BM25F|HMM|speech recognition correction|sentence correction|keywordsdomain document

要約(日本語): In this paper we proposed a method for correction of non-native English speakers' speech, based on the speech's domain focused document.Our approach consists of three important steps. Detailed document analysis in the preprocessing phase, finding the document part corresponding to recognized by SR sentence, and sentence correction part.The analysis creates the BM25F method lemmas' vectors, and the keyword vectors.Finding the corresponding part is realised using the HMM based method.The correction is conducted by calculating the final probability connecting keywords comparison, BM25F lemmas comparison and HMM based method, for each candidate sentence, and choosing the most probable.The candidate sentences are obtained from the result of speech recognition software.

PDFファイルサイズ: 245 Kバイト