Transformation of Japanese Classical Sentence to Japanese Modern Sentence and Acquisition of Syntactical Data on Classical Words

We tried a method of translation from a classical Japanese language to modern Japanese. These languages have many similar points in syntax and vocabulary, but not the same. For example, they have syntactically similar word ordering in a sentence and a similar part-of speech set with slightly different conjugation system. For modern Japanese, we have word dictionaries with syntax data such as case frame data of verbs and concept data of nouns. For classical Japanese, we only have classical language dictionaries composed of a sequence of records with a pair of classical Japanese head word and corresponding modern Japanese words. In view of these points, we tried a simple translation method that is word-by-word translation at first. However, if the source word has many meanings, we get plural object words. Then, we try to select suitable modern Japanese words using case frame data and concepts data of modern Japanese. We produced a translation program and performed an experiment of translation from sentences of "Ise Monogatari" (in the tenth century) to modern Japanese. As a result, our method is useful for the classical words which have corresponding modern Japanese single words defined in modern Japanese syntax dictionaries. However, in case of the classical word which have no corresponding modern Japanese single word, and in case of modern Japanese words which have not enough dictionary data, our method fails to get appropriate translation. We need to improve and enrich modern Japanese dictionaries or to try to use "valence" data in stead of case frame data. It is desirable that parallel translation corpora, available to computer software, are produced. These are future studies.