Ontological approach applied to the C99 algorithm

C99 is one of the most famous and popular algorithms in the literature. It was designed for the detection of the topic boundaries in textual documents. C99 like many other algorithms is based on what is called lexical cohesion analysis.

In this paper, we use this algorithm to locate the theme changes in a specialized text. How-ever, we introduced the elements of domain ontology in order to exploit the semantic linking between terms that appear in such text.

Indeed, we question the pre processing operation used by C99. It performs stemming process by the suffix stripping of the words in order to generate finally the word frequencies matrix; this way of doing generates significant noise which decreases the algorithm performance. We suggest as an alternative to use the ontological analysis process that we have imple-mented in a previous work.

Finally, we evaluate the effectiveness of this option versus the basic version of C99 by using a specific text corpus formed by concatenated sections dealing with different topics of the domain.