The Comparative Research on the Segmentation Strategies of Tibetan Bounded-Variant Forms

Abstract

The segmentation of Tibetan bounded-variant forms (TBVFS) is one of the most foundational tasks in text processing and the segmenting results directly influence the word segmentation, portaging, syntactic parsing and the Named Entity Extraction and so on. At present, the segmenting results are unsatisfactory and cannot be applied in practice. In this article, authors firstly describe the features of TBVFS, their distributions and then test the segmenting results by using two different segmentation strategies and conclude that Statistics-based methods for morpheme position tagging is better than Rule-based methods. If some rules are used to adjust a part of mistaken segmentations in the post processing, this kind of segmentation problem can be resolved.