To provide a speech recognition system capable of speech input using a dictionary for keyboard input without making a language model complicated.
A language model attached to a speech recognizing means 22 does not determine a word to be converted into Kanji(Chinese character) by using a plurality of word groups corresponding to a pattern of speech information like a conventional language model, but converts it into a phonetic character string such as Hiragana(cursive form of Japanese character) or Roman characters. A converting means 26 converts the phonetic character string which is thus converted is converted into Kanji, phrase by phrase, according to the dictionary 28. The conversion result is displayed by a display means 30. Consequently, the same dictionary can be used for speech input and keyboard input and the constitution of the speech recognizing means 22 (language model) can be simplified.
IKETA YOSHIKAZU
KITAMURA MINORU
TOMOYOSE KATSUAKI