Building hypertext links by computing semantic similarity

Most current automatic hypertext generation systems rely on term repetition to calculate the relatedness of two documents. There are well-recognized p roblems with such approaches, most notably, a vulnerability to the effects of synonymy (many words for the same concept) and polysemy (many concepts f or the same word). We propose a novel method for automatic hypertext genera tion that is based on a technique called lexical chaining, a method for dis covering sequences of related words in a text. This method uses a move gene ral notion of document relatedness, and attempts to take into account the e ffects of synonymy and polysemy. We also present the results of an empirica l study designed to test this method in the context of a question answering task from a database of newspaper articles.