基于重排序融合的社会图书检索系统

用图书的出版信息和用户生成的社会信息从社会媒体中搜索出相关的图书已成为信息检索系统的一个研究热点。然而大部分的信息检索系统都是由单一的检索方法构成，随着用户需求的不断增加，这些系统难以满足用户需求。针对上述问题，提出了一种基于重排序融合的图书检索系统。首先，使用伪相关反馈技术对用户查询内容进行扩展，并将检索结果作为初排序结果；其次，使用用户生成的社会信息特征对初排序结果进行重排序；最后，采用排序学习模型对多种重排序策略得到的结果进行融合。在INEX 2012-2014 Social Book Search公开数据集上针对其它先进检索系统进行了对比实验，实验结果表明，系统的性能（NDCG@10）优于其它方法构成的图书检索系统。

An integrated system for social book search

Searching and navigating books with professional and user-generated content from social media is a hot topic in information retrieval systems. However, most methods are specially designed as a purely searching system. With the increasing demands, these systems have difficulties in satisfying users. Aiming at these problems above, this paper presented an integrated social book search system based on several retrieval techniques. The system firstly enriched the queries with pseudo-relevance feedback to improve the initial ranking results. Then several re-ranking models with different features re-ranked the improved results as their input. Finally, the system structured a learning-to-rank model to combine several re-ranking results. Experimental results on the INEX 2012-2014 Social Book Search query dataset show that the proposed system can achieve better performance(NDCG@10) compared to state-of-the-art search system.