关键词: 用户意图, 依存关系, 动词聚类, 潜在狄雷克来分配模型

Abstract:

In order to understand users’goals in web searching better, an approach which makes use of hierarchical dirichlet process (HDP) model to cluster the verbs of users’goals is proposed. Verbs are good indications of users’goal, and good verb clustering therefore causes good clustering of users’goals. Verbs of users’goals are represented by cooccurrence nouns and nouns that have dependency relations with the verb. Experiments show that HDP has better performance in verbs clustering than latent Dirichlet allocation and dirichlet process mixture model.

Key words: user goals, dependency relation, verb clustering, latent Dirichlet allocation