AutoTagTCG : A Framework for Automatic Thai CG Tagging

Recently, categorical grammar has been focused as a powerful grammar. This paper aims to develop a framework for automatic CG tagging for Thai. We investigated two main algorithms, CRF and Statistical alignment model based on information theory (SAM). We found that SAM gives the best results both in word level and sentence level. We got the accuracy 89.25% in word level and 82.49% in sentence level. SAM is better than CRF in known word. On the other hand, CRF is better than SAM when we applied for unknown word. Combining both methods can be suited for both known and unknown word.

• Type: Conference
• Year: 2010
• Where: LREC
• Authors: Thepchai Supnithi, Taneth Ruangrajitpakorn, Kanokorn Trakultaweekool, Peerachet Porkaew
• Keywords: Education, LREC 2010, Sentence Level, Unknown Word, Word Level