PURPOSE: To facilitate the processing of KANA-KANJI conversion, by decomposing an input sentence into paragraphs centering on independent words to supply them in the form of KANA letters, decomposing those paragraphs into words to display the conversion candidate characters from the head of order of them and then deciding successively the candidate characters intended by an operator.

CONSTITUTION: An input sentence is decomposed into paragraphs centering on independent words and supplied to an input preprocessing part 1 in the form of a KANA sentence by means of a control key. The information on English letters, numeric characters, paragraphs, etc. are recognized, and a KANA letter string to be controlled in extracted and transferred to a word extracting part 2. Then a dictionary is retrieved to extract the candidates of word strings that can be allowed grammatically. A homophone discriminating part 3 decides the most prominent candidate from the frequency information in case plural word string candidates exist. An output controlling part 4 displays corresponding characters for each paragraph. When receiving a control signal indicating the next candidate, the part 1 displays the second candidate in place of the first most prominent candidate. In such a way, the KANA-KANJI conversion processing is facilitated.