情境搜索的关键技术研究

发布时间：2018-07-08 12:04

本文选题：情境搜索 + 个性化搜索　；参考：《北京邮电大学》2013年硕士论文

【摘要】：传统的互联网搜索在搜索信息结果的完备上已经取得很好的效果,但是很难满足用户的个性化需求,无法得到因人而异的搜索结果,无法根据用户所在情境搜索出满足情境的个性化信息。本文对情境搜索的关键技术做了研究。在调研了国内外情境搜索的研究现状后,发现大部分的情境搜索只是单一的将位置和时间信息加入到个性化服务中,所以提供的搜索结果的个性化程度并不高。如果能够将用户的社交情境加入到搜索引擎中,并结合多方面的个性化技术,势必带来更具个性化的搜索服务。以此为出发点,本文从搜索引擎的架构和个性化服务的提出入手,调研了社交网络分析、扩展查询、情境获取以及个性化推荐的相关技术,给出了基于移动平台的情境搜索个性化融合建模的构思。从主体、客体、主体关系三个角度为搜索建模,以客体黄页特征模型和历史行为为依据建立用户兴趣模型、以社交关系为参考建立用户社交图谱。在社交图谱的基础上为增加搜索结果的新颖性和可信度,使用基于邻域的个性化推荐算法相关技术,增加了来自社交推荐的部分搜索结果。在以上的建模基础上,进行线性加权融合,构成情境搜索的融合模型。然后,本文设计了一个用于移动黄页平台的情境搜索系统,在系统的设计中将融合模型算法应用到系统设计中,并提出评价情境搜索系统的几个关键指标。在随后的实验数据中,从用户调查、在线测试以及离线评测三方面对3000个用户和46万条相关黄页做了大量的测试,论证了融合模型算法在情境搜索系统中的可行性。最后,对本文的研究成果作了总结,提出了当前工作中的不足和下一步应当进行的研究。
[Abstract]:Traditional Internet search has achieved good results on the completion of search information results, but it is difficult to meet the personalized needs of users, and can not get individual search results. It is impossible to search the personalized information to satisfy the situation according to the user's situation. This paper studies the key technologies of situational search. After investigating the current situation of situational search at home and abroad, it is found that most of the situational search only adds location and time information to the personalized service, so the personalized degree of the search results provided is not high. If the user's social situation can be added to the search engine, and many aspects of personalized technology can be combined, it will bring more personalized search service. Starting with the structure of search engine and the proposal of personalized service, this paper investigates the related technologies of social network analysis, extended query, situation acquisition and personalized recommendation. The concept of personalized fusion modeling of situational search based on mobile platform is presented. Based on the object yellow page feature model and historical behavior, the user interest model is established, and the user social map is established based on the social relationship. In order to increase the novelty and reliability of search results, the neighborhood based personalized recommendation algorithm is used to increase the search results from social recommendation. On the basis of the above modeling, linear weighted fusion is carried out to form the fusion model of situational search. Then, this paper designs a situational search system for mobile yellow page platform. In the design of the system, the fusion model algorithm is applied to the system design, and several key indicators for evaluating the situational search system are put forward. In the subsequent experiment data, 3000 users and 460000 related yellow pages were tested from three aspects of user survey, online test and off-line evaluation, which demonstrated the feasibility of the fusion model algorithm in the context search system. Finally, the research results of this paper are summarized, the shortcomings of the current work and the next step should be studied.
【学位授予单位】：北京邮电大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TP391.3

【参考文献】