On the Generality of Thesaurally derived Lexical Links
Jeremy Ellman, John Irving Tait · 2000
Cohesion is that property of a text that allows it to be read as a unified entity rather than a series of unconnected sentences. Lexical cohesion may be detected using an external thesaurus and the resulting representation used in a variety of language processing tasks. Our particular interest is in determining whether texts of different genres are similar in meaning. For this, we wish to derive a measure based on lexical cohesion. Consequently, we need to determine if lexical cohesion is independent of genre or a function of it. This paper examines the statistics of lexical cohesive relations. Our method involves determining the distribution of lexically cohesive relations in several book length texts. These a re shown to have different reading complexities, but equivalent cohesive properties. From this, we conclude that lexical cohesion is independent of reading complexity.