Use of linguistic features in context-sensitive text classification

Abstract

Many popular Text Classification (TC) models use simple occurrence or words in a document as features to base their classifications. They commonly assume word occurrences to be statistically independent in their design. Although such assumption does not hold in general. these TC models arc robust and efficient in their task. Some recent studies have shown contextsensitive TC approaches were able to perform better in general. On the other hand, although complex linguistic or semantic features may intuitively be more relevant in TC. studies on their effectiveness have produced mixed and inconclusive results. In this pajier. we present out investigation on the use of some complex linguistic features with two context-sensitive TC methods. Our experimental results show potential advantages of such approach. © Springer-Verlag Berlin Heidelberg 2006.