Abstract: In the light of the differences between professional and universal search engines, a novel Query Recommendation algorithm based on word Co-occurrence and HITS algorithm（QR-CH） is proposed for professional search engines. To improve the relevance between the recommended words and the initial query, QR-CH utilizes the HITS algorithm to order the candidates which are filtered by word co-occurrence, and then chooses the candidates with high relevance as recommended words. At the same time, the algorithm reduces the redundancy effectively. Whether the recommended word is redundant depends on query results, which are also ordered by the HITS algorithm. QR-CH stores the recommended words in a domain knowledge tree which is used for query recommendation. The results of the experiment show that QR-CH is superior to the existent similar algorithms in both the relevance and the redundancy.

Key words: query recommendation, word co-occurrence, Hypertext Induced Topic Search（HITS） algorithm, professional search engine

摘要： 根据专业搜索引擎的特点，提出了一种新颖的基于词语共现与HITS算法的查询推荐算法QR-CH（Query Recommendation algorithm based on word Co-occurrence and HITS algorithm）。该算法一方面利用HITS算法对基于词语共现筛选出的关联词按语义关联性进行排序，选取排序靠前的关联词作为推荐词，提高了推荐词与原查询词的相关性；另一方面使用HITS算法排序关联文档，从查询结果文档集的角度来判断推荐是否冗余，降低了推荐词的冗余性。该算法将推荐相关的信息存储到知识树中，利用知识树实现查询推荐。实验结果表明QR-CH算法在推荐词的相关性和冗余词的判断方面均优于文献中已有的类似算法。

关键词: 查询推荐, 词语共现, 超链诱导主题搜索（HITS）算法, 专业搜索引擎