基于社会信息重排序的图书检索方法

随着社交网络的兴起和发展，互联网上出现了大量与商品有关的社会信息。如何利用这些社会信息结合商品元数据进行检索和推荐是信息检索领域中一个热门的研究问题。本文以社会图书检索为例，提出了一种通用的信息检索方法来解决这一问题。首先，通过分析原始图书数据集和图书的用户标签、用户评分和流行度等社会信息，从图书中提取不同的社会特征构建特征矩阵；然后分别计算图书在各种社会特征上的相似度，并使用不同的策略对搜索引擎返回的排序结果进行重排序；最后使用学习排序的方法进行重排结果融合，得到最终的图书检索结果。在实验中，使用该检索方法在INEX Social Book Search 2015和2016数据集上分别进行了训练和测试。结果表明，相比现有的技术，该检索方法能够有效提升图书检索的效果。

A Book Retrieval Method Based on Social Information Reranking

With the development of social networks, a great deal of social information arised on the internet. How to make full use of these social information and products metadata in search and recommendation is a hot topic in the research field of information retrieval. In this paper, the authors took social book search as an example, proposed a general search-recommendation hybrid system for this problem. Firstly, by analyzing the original book data set and the social information of books, such as tags, rating and popularity, this study constructed the feature matrix by extract social features in books. Afterwards, we computed the similarity of books in different social features, and reranked the initial result of search engine with different strategies. Finally, we used learning-to-rank technique to combine a wide range of diverse reranking results. In this experiment, we trained and test the proposed system on the INEX Social Book Search 2015 and 2016 datasets with this method respectively. The result showed that our system can effectively improve the retrieval performance compared with other state-of-the-art systems.