XML图书搜索中基于用户社会关系的查询推荐方法

QUERY RECOMMENDATION METHOD IN XML BOOK SEARCH BASED ON USERS' SOCIAL RELATIONS

基于关键词的图书搜索系统由于用户输入查询词的模糊性和简单性往往需要利用查询推荐技术对查询词进行扩展。目前的图书查询推荐方法不能辨别出不同用户在不同时期的图书请求意图和兴趣。提出一种基于用户社会关系的查询推荐方法,首先通过分析用户个人资料信息建立用户社会关系对象集合;其次获取用户社会关系对象对图书的标记词,计算输入查询词与标记词之间的共现率并建立用户社会关系标记词推荐集合,选取与查询词共现率最高的标注词进行查询词扩展。在实际图书数据集上的实验表明,该方法大大提高了查询结果的NDCG@10值,提高用户的满意度,表明该方法具有可行性。

Since the query words inputted by the user are fuzzy and simple, the keyword-based book search system usually requires using query recommendation technology to expand the querying words. So far the book search query recommendation methods could not distinguish different request intentions and the interests of different users in different periods. This paper presents a query recommendation method, which is based on users＇ social relations. First, it creates a collection of objects in regard to users＇ social relations by analysing the information of users profiles; Secondly, it obtains the book mark words marked by users social relations objects, and then calculates the rate of co-occur- rence between the query word inputted and the mark words, and establishes the collection of the recommended users social relations mark words; At last, it selects the mark words with the highest rate of co-occurrence as the words for query expansion. Experiment on the actual book dataset shows that this method greatly improves the nDCG@ 10 values of the query results and users＇ satisfaction, this means that the method is feasible.