The development of Tagged Uyghur Corpus

The development of tagged Uyghur corpus Yusup Aibaidula Department of Computer Science of Physi- Math, Information Xinjiang Normal University, Urumqi, Xinjiang, China E-mail: yusup2002@sohu.com Kim-Teng Lua* School of Computing, National University of Singapore, 3 Science Drive 2, Singapore 119260 luakt@comp.nus.edu.sg Abstract: The history and development of Uyghur language is introduced. After a brief introduction to the development of Uyghur words, morphology and syntax, we explain our developing of a computer-aided contemporary Uyghur language tagging system. The coverage of this corpus, the resources building, the rules for syncopating and tagging etyma and termination, and the tagging of a corpus using a small tagset are explained. Some practical methods solving problems in Uyghur language tagging are also proposed. Key word: history and developmnet of Uyghur language, Uyghur tagged corpus, Uyghur language tagging system. 1 Introduction Processing raw corpuses is fundamental to the research of the corpus linguistics. Since the first corpus was built in USA in 1970's, many other corpuses had been developed world wide. In the early days, a corpus usually contains only one million words and only text corpuses are collected. From 1980's, research in linguistic corpuses becomes very active. Some major projects at that time are: the Lancaster Treebank, the tagged LOB Corpus and the Birmingham Corpus. In the 1980's the corpuses are large and there are more varieties. Before 1990s, almost all are English corpuses. From 1986 to 1994, several different part-of-speech (POS) automatic tagging tools such as CLAWS1, CLAWS2, CLAWS3 and CLAWS4 are developed. At the same time, many progresses had been made in the corpus syntax analysis and tagging. The PennTreebank, developed by the Pennsylvania University of USA combined and integrated many corpus processing tools such as: the Church POS tagging tools, the Fidditch syntax analysis kit of the Hindle system and etc. These tools greatly enhanced the efficiency of the corpus tagging. After 1990s, many non-English corpuses are developed. These include the Japanese ER corpus and the news manuscript corpus of NHK. In the past ten years, much progress has been made in the development of Chinese corpuses. These include: the automatic syncopating ([LNY87], [XHS91]), POS tagging ([BXH92], [BXH92]), dependency tagging ([ZH94], [LZH93]) and etc. From 1992, researchers in the Institute of Computational Linguistics in Peking University have been working in the integrated processing of the Chinese corpuses. They have developed many new technologies and tools for the Chinese corpus processing. From the viewpoint of the sources of a language, a corpus can be divided into text or speech, spoken language or written language and single language or multi languages. 228 So far, many tagsets have been developed for POS tagging, phrase tagging, dependency relation tagging, case relationship, syntax tree, semantic relationship and so on. As much as we know, the beginning of building Uyghur corpus is in 2001. So there are far more works to be done for this language. 2 Tagging of the contemporary Uyghur corpus The development of Uyghur alphabets has a long history. Uyghur means "combination" or "solidarity". Before the 9th …

Content truncated to 500 words