| 摘要 |
PURPOSE:To avoid invalidation of all inputs and to eliminate the necessity of rein put even in case the analysis of the KANA (Japanese syllabary)/KANJI (Chinese character) input data fails partly, by storing a wrong paragraph end word if decided to a paragraph candidate store part as a wrong word via an unfixed word generating means. CONSTITUTION:Only the KANA data are stored in a KANA character sentence reception/storage part 2, and a Japanese dictionary retrieving means 4 stores the paragraph candidates to a paragraph candidate storage part 7 through a paragraph candidate development means 3 based on the 1st KANA digit position. The end digit position of a paragraph end candidate is also stored and '1' is added to the KANA digit position at a base point. If no digit position equal to the paragraph end exists at and after the base point, an unfixed word generating means 6 produces an unfixed word paragraph of a digit. This digit is used as a paragraph to develop the paragraph candidates which are stored in a paragraph candidate buffer 8. The grammar error data sets up an error flag to the data in the buffer 8. A paragraph candidate re-editing means 11 deletes the unfixed word data as an error together with the data decided as an error by a document candidate junction inspection means 9 against the data of the buffer 8 as long as an overall connection state is secured with no unfixed word existing. Then the effective data is re-edited to the buffer 8. |
