フォートラン・プログラムの音声認識システム

概要

論文の詳細を見る
Speech recognition system of programs in JIS Fortran 3000 level uttered with a rather longer pause between successive words has been constructed. It consists of an acoustic analyser and a linguistic processor, and has an automatic error correcting function. In linguistic processor two types of lexicon are prepared. One of them is concerning to gross features. The other consists of the phoneme string of each word (total 79 words), which is essentially spelled in Roman letters and also some rewriting rules of phonemes are contaiuned in each word. By syntatic and semantic analyses and first type of lexicon, some possible input words are predicted. The phoneme sequences of predicted words are matched with the input one and the word having the highest reliability is adopted as the output. Each statement of programs is syntactically examined and is automatically corrected whenever there is any syntactic error. If some confusions among English letters are allowed, we can recognize any program.
一般社団法人情報処理学会の論文
1977-05-15