Role Assignment to Thai Word Using Multilayer Perception Network

Thai language has some certain characteristics, which lead to a number of interesting research problems in natural language processing. For instance, Thai words in a sentence do not have feature embedded form, i.e., Thai words always have the same form in any sentences. This partially explains the lack of feature information in Thai part of speech (POS). Thai POS tagging only tags word category, such as noun, pronoun, verb, proposition and so on. Category information alone is inadequate for further language processing. This paper proposes the idea of tagging word role in a sentence using multilayer perceptron as a toll. The role assignment information can be used in complimentary to category assignment in Thai language processing.

Keywords:  Part of speech tagging, Thai language processing, POS tagging with neural network, multilayer perceptron network, role feature tagging

Corresponding author: E-mail: ponrudee@it.kmitl.ac.th