【正文】
? Run and (WFST decode for triphone) ? Run (Viterbi for mono) ? Run (Viterbi for triphone) ToDo (Opt.) ? Train LM : Use YOUR training text or even YOUR lexicon. ? Train LM (ngramcount) : Try different arguments. ? ? Watch online courses on coursera (Week2 LM) ? ? Read 數(shù)位語音處理概論 ? (Viterbi) ? (Language Model) ? (WFST) ? Try different AM/LM binations and report the recognition results. Questions ? 。 L。 B means, C is A posed with B WFST Component ? HCLG = H。 f 1 plement $train_text /exp/lm/ Language Model : ngramcount (1/3) ? /share/srilm/bin/i686m64/ngramcount ? order 2 (You can modify it from 1~3) ? kndiscount (modified KneserNey smoothing) ? text /exp/lm/ (Your training data file name on ) ? vocab $lexicon (Lexicon, as shown on ) ? unk (Build open vocabulary language model) ? lm $lm_ou