文本信息检索质量评估技术发展趋势及展望

Development Trends and Prospects of Research on Quality Evaluation Techniques for Text Information Retrieval

我国超过9.97亿网民通过输入文字方式进行信息检索,约94.1%的用户表示文本信息检索返回的结果中包含用户不期望的信息,约28.5%用户表示对文本信息检索返回的结果不信任。本文首先对国内外关于文本信息检索质量评估技术研究现状进行分析,现有的文本信息检索结果评价研究主要基于假设查询词是用户搜索需求的映射,采用多样化的检索结果技术尽可能地满足不同用户的信息搜索需求。然后,针对用户个性化检索意图对检索结果的可信性判断问题,分别从提升信息检索质量评估精度、增强用户检索体验和加强用户隐私保护等3个维度,给出文本信息检索质量评估技术发展所面临的主要挑战。在此基础之上,提出面向用户体验增强的信息检索动态评估模型构建、基于卷积神经网络的信息质量评估方法和面向用户隐私保护的检索意图识别方法等3个主要方面的技术研究发展趋势及对未来的展望。

More than 997 million Chinese netizens search for information by inputting text into search engines,94.1%of users reported that the search engine returned results containing information that was not expected by the user,and 28.5%of users expressed distrust of the search results provided by the search engine.This article first analyzes the current research status of text information retrieval quality evaluation technology at home and abroad,the existing research on the evaluation of text information retrieval results is mainly based on the assumption that the query term is the mapping of user search needs,using diversified search results technology to meet the information search needs of different users as much as possible.Then,the personalized retrieval intention of users greatly affects the credibility judgment of retrieval results,main challenges faced by the development of text information retrieval quality evaluation technology are presented from three dimensions:improving the accuracy of information retrieval quality evaluation,enhancing user retrieval experience,and strengthening user privacy protection.Based on these,three main technological research trends and research ideas are proposed,including the construction of a dynamic evaluation model for user experience enhancement in information retrieval,information quality evaluation methods based on convolutional neural networks,and retrieval intent recognition methods for user privacy protection.