Uyghur vowel weakening processing system

Abstract

Uyghur vowel weakening has three kinds of phenomena, such as vowel dropping, inserting and neutralizing. When words added some suffixes, there always occurs some weakening take place. In Uyghur, vowel neutralizing is very common, and when we need to find the stem form of an inflected stem, it is hard to identify the neutralized vowel's original vowel without a dictionary in some cases. In Uyghur, there are eight vowels. But disregarding vowel length distinction, one identifies nine vowels in Uyghur. Actually, the neutral vowel in Uyghur represents two distinct phonemes, even though they share the same set of allophones and are orthographically alike. In this paper, we introduce our vowel weakening processing system and the Maximum Entropy (MaxEnt) based neutral vowel phoneme identification model.