Effects of Mora Phonemes on Japanese Word Accent

Effects of Mora Phonemes on Japanese Word Accent Yasuyo TOKUHIRO Graduate School of Japanese Applied Linguistics, Waseda University 1-7-14, Nishi-Waseda, Shinjuku-ku Tokyo, 169-8050 Japan tokuhiro@toki.waseda.jp Shizuo HIKI Waseda University 2-579-15 Mikajima Tokorozawa, 359-1192 Japan hiki@waseda.jp Abstract Effects of mora phonemes on Japanese word accent was analyzed statistically, utilizing a set of about 124,000 frequently used common nouns derived from the Japanese Word Dictionary edited by EDR (the Japan Electronic Dictionary Research Institute, Ltd., Japan). In this analysis, Japanese syllable was defined as preceding consonant (+semi-vowel) +following vowel, accompanied / not accompanied by mora phoneme. Mora phonemes adopted here were elongated portion of vowel: /H/, chocked sound: /Q/, syllabic nasal: /N/, and /i/ portion of diphthongs: /I/. The average ratio of attachment of accent to vowel type syllables was 13% and to semi-vowel +vowel type syllables 16%. In both types, the difference according to the kind of following vowels was not significant. As for the effect of mora phonemes, the ratios of accent attached to the syllables accompanied by mora phonemes were about several percents higher than their average for any kind of following semi-vowels and vowels. The average ratio of accent attached to vowel and semi-vowel +vowel type syllables was 14%, but in syllables of any kind of preceding consonants accompanied by mora phoneme, the ratios were more than several percents higher than the average. The words which involve the mora phonemes were categorized by each kind of mora phoneme, and compiled into the list in the order of Japanese syllabary, and the lists of different number of syllables composing each word in the order of Japanese syllabary. In addition, the list for retrieving the possible candidate words which could be replaced if the mora phonemes were missing through imperfect hearing or pronouncing. Those databases are useful for language teaching and speech training, as well as for basic research on speech processing. 1 Introduction From the Japanese Word Dictionary edited by EDR (Japan Electronic Dictionary Research Institute, Ltd., Japan), about 300,000 common nouns were extracted, and the foreign words which occupied 15% of them were deleted. About 80% of the inventory of 260,000 remaining words belonged to the groups of the words having the same pronunciation (homonyms), as many of the words were listed separately according to the precise differences in meanings or notations in the “Headwords” of the EDR dictionary. So, the groups of homonyms in the inventory were sub-divided step by step into the groups having the same “Concept Identifier” (the minimum unit of the concept of word), “Headconcept” (finer unit than the head words), then, “Headword”, and a representative word for each of the sub-groups were extracted, based on higher “Concept Occurrence Frequency,” in each of the steps. In case that the "Concept Occurrence Frequency" was even, a word which contains larger number of Chinese character was selected. By defining homonyms step by step wider in this way, the original word inventory was consolidated into the resulting set of about 124,000 words. This corresponded to about 50% …

Content truncated to 500 words