Abstract: Words mismatch between queries words and documents lead to a number of related doucuments can not be successfully retrieved in information retrieval，which affects the effectiveness of retrieval results.This paper proposes and realizes a model based on the conceptual graphs and HowNet，and proposes a kind of relevance feedback algorithm，which achieves the similarity of new documents and query words with re-weighting word items by vector space model，and proposes semantic retrieval model.Experiments show that the new idea proves this method is effective.

Key words: information retrieval, similarity, vector space model, HowNet, relevance feedback

摘要： 由于查询与文档中词语的不匹配现象导致一些相关的文档不能被成功地检索出来，在信息检索的研究与实现中，这是影响检索效果的一个很关键的问题。把概念图和知网结合起来，提出对应的相关反馈算法，重新计算词项权重，利用向量空间模型和语义相似度进行语义检索，并给出了语义检索模型。实验结果显示该方法取得了良好的效果。

关键词: 信息检索, 相似度, 向量空间模型, 知网, 相关反馈