| 摘要 |
PURPOSE:To obtain the language processor for recognition which is used for speech and character recognition and performs language processing at a high speed. CONSTITUTION:A lattice generation part 3 generates a syllable lattice from an inputted syllable candidate. A reduced writing conversion part 4 converts the writing of each syllable candidate constituting the syllable lattice into a reduced writing to obtain a reduced writing syllable lattice. A reduced writing language processing part 5 expands the reduced writing syllable lattice and performs reduced writing language processing by using a reduced writing dictionary 6 to obtain a reduced writing paragraph candidate. A candidate deletion part 7 deletes a syllable candidate which is not considered to exist from the syllable lattice by referring to the reduced writing paragraph candidate. An original writing language processing part 8 expands the syllable lattice which is decreased in the number of constituent elements and performs original writing language processing by using an original writing dictionary 9. Thus, the language processing is performed at a high speed by the reduced writing language processing which uses reduced writings small in the number at categories and the language processing using the syllable lattice which is small in the number of constituent elements. |
