Аннотация
We introduce two different approaches for clustering semantically similarwords. We accommodate ambiguity by allowing a word to belong to severalclusters.Both methods use a graph-theoretic representation of words and theirparadigmatic relationships. The first approach is based on the concept ofcurvature and divides the word graph into classes of similar words by removingwords of low curvature which connect several dispersed clusters.The second method, instead of clustering the nodes, clusters the links in ourgraph. These contain more specific contextual information than nodesrepresenting just words. In so doing, we naturally accommodate ambiguity byallowing multiple class membership.Both methods are evaluated on a lexical acquisition task, using clustering toadd nouns to the WordNet taxonomy. The most effective method is linkclustering.
Пользователи данного ресурса
Пожалуйста,
войдите в систему, чтобы принять участие в дискуссии (добавить собственные рецензию, или комментарий)