古典中國語テキストの知識處理について

Abstract

Chinese characters have a features of morpheme, each character represents phonetic values and grammatical and/or semantical values by its shape. Shape can be regarded as spelling in alphabetical scripts. These characteristics of Chinese character mean that character level phenomena and higher level phenomena are observed as mixed behavior of character and we often don't distinguish them. In addition, shape (graphical information) and phonetic value also have such kind of structures. Each modal value has physical level (e. g. written shape, spoken pronunciation) and abstract level (e. g. abstract glyph, phoneme), and we can find various intermediate abstract/concrete levels, so we can draw a gradation of various levels. Each modal value has a gradation of abstract/concrete objects, so each character has a multidimensional gradation. In the present text processing mechanism for classical Chinese, however, characters are basically separated by abstract character layer and glyph (and glyph image) layer(s), and character level and higher level are separated by isolated layers, so they are not seamless. In addition, grammatical and semantical processing are very poor. Some difficulties of classical Chinese processing may be caused by the framework of text processing. Anyway, the current situation requires a lot of human resources to develop complicated markup texts and/or knowledge resources for semantical fields and maintain increasing data. In addition, domain specific data may not produce other kinds of data. Tools chains are very poor or not available. It may be a serious problem of the progress of classical Chinese informatics. This paper describes a feasible knowledge processing methods for classical Chinese text to resolve such kind of situations. It mainly focus morphological analyzer, glyph corpora and character ontology, then we discuss about multimodal knowledge processing mechanism based on their integration.