Abstract: Bayesian network retrieval models are suitable models to deal with information retrieval problem because they are appropriate tools to store the conditional probabilities and semantic meanings among terms and compute the similarity between user’s query and documents.In this paper the simple Bayesian network retrieval model and the extended models are evaluated detailed with the Chinese collection，experimental results show that the extended models behave better than others，realizing semantic retrieval to some extent.

Key words: Bayesian network retrieval models, term similarity, information retrieval, synonyms

摘要： 贝叶斯网络检索模型可以表示术语间的条件概率和概念语义，并依此预测用户查询和文档间的相似度，是解决信息检索的有效手段。通过构造中文测试集合，对简单贝叶斯网络检索模型和扩展的贝叶斯网络检索模型的性能进行详细评估，实验证明扩展模型可以有效地提高检索性能，在一定程度上实现了基于语义的信息检索。

关键词: 贝叶斯网络检索模型, 术语相似度, 信息检索, 同义词