Thai Morphological Analyses Based on the Syllable Formation Rules

概要

論文の詳細を見る
The Thai syllable formation rules were deduced from an orthographical analysis of Thai. At the morphological level, segmentation was analyzed by the ordinary longest-match method for the input of Thai text (the Law of Three Seals: 20,631 sentences), a revised method of segmentation, called the Syllable Longest-Match method (SLM), which incorporated a mechanism of back-tracking for each phoneme based on the syllable formation rules when the segmentation failed, was then devised to reduce the number of unsuccessful cases. This method indicated that the ratio of segmentation is a 98.0%, which is 2.8o/o greater than the ordinary method in terms of sentences. A finite automaton model which employs the automatic segmentation from a sentence into monosyllables without reference to a dictionary, called Thai syllable recognizer, was also proposed. A revised Thai syllable recognizer was also devised, in which knowledge rules based on the heuristics derived from the analysis of unsuccessful cases were adapted the existing syllable formation rules. This gave a ratio of segmentation is 93.9% in terms of sentences for the input of same text.
一般社団法人情報処理学会の論文
1992-12-31