This paper proposes the use of a hybrid
approach integrating between rule-based method and statistical
approach using contextual behaviour features reflecting natural
language behaviour for Thai sentence-breaking. The
performance of Thai sentence-breaking using a number of words
in a chunk, existence of verb, and rules are compared.
Experimental results show that using a number of words in a
chunk achieves higher accuracy than other features. Moreover,
integration of those features and rule-based method achieves
better accuracy. The space-correct and false-break scores are
93.54% and 2.99% respectively.