AUTOMATIC IDENTIFICATION OF COHESION IN TEXTS: EXPLOITING THE LEXICAL ORGANISATION OF ROGET'S THESAURUS

The identification of semantic relations between words can be applied to the subsequent identification of cohesion in texts. Groups of cohesive portions of text can be used to identify document structure and sub-topic areas. One method of automatically identifying the semantic relations between words is the utilisation of an existing lexical knowledge source, such as Roget's Thesaurus. A technique has been developed that exploits the lexical organisation of the thesaurus and this has been applied to the identification of semantically related words and of cohesion in texts. The development of this technique is outlined and the results from experiments conducted to investigate the application of the technique are presented and discussed.