Study on automatic extraction method of Tibetan new words

This paper proposes a model to automatically extract Tibetan new words. Through building the dynamic Tibetan corpus from 2009 to 2012, which covers more than 18 Tibetan network media of Tibet, Qinghai, Sichuan, Gansu and Yunnan, we research on the key techniques of Tibetan new word extraction: (1) using statistical method to establish Tibetan new words knowledge base; (2) using information entropy and vector space module similarity calculation to extract/filter Tibetan new valid words; (3) using word co-occurrence techniques to extract Tibetan new meaning words.