Deriving concept hierarchies from text

摘要: Abstract : This paper presents a means of automatically deriving hierarchical organization concepts from set documents without use training data or standard clustering techniques. Instead, salient words and phrases extracted the are organized hierarchically using type co-occurrence known as subsumption. The resulting structure is displayed series menus. When generated retrieved documents, user browsing menus provided with detailed overview their content in manner distinct existing summarization methods used to build simple, but appear be effective: smallscale study reveals that hierarchy possesses properties expected such general terms placed at top levels leading related more specific below. formation presentation described along some other informal evaluations. into concept derived itself undoubtedly one goal information retrieval. Were this achieved, would form somewhat like manually constructed subject hierarchies, Library Congress categories, Dewey Decimal system. only difference being categories customized itself. For example, collection media articles, category "Entertainment" might near level; below it, (amongst others) find "Movies", entertainment; that, there could "Actors & Actresses", an aspect movies. As can seen, arrangement provides topic those articles.