基于深度学习挖掘用户搜索主题研究  

Research on Mining User Search Topic Based on Deep Learning

作　　者：宋毅[1] SONG Yi(School of Data Science and Artificial Intelligence,Harbin Huade University,Harbin 150025,China)

机构地区：[1]哈尔滨华德学院数据科学与人工智能学院,黑龙江哈尔滨150025

出　　处：《计算机技术与发展》2021年第1期43-47,共5页Computer Technology and Development

基　　金：国家自然科学基金(61672185);黑龙江省自然基金(F2015046);黑龙江省教育科学规划重点课题(GJB1320081)

摘　　要：主要研究了基于深度学习技术挖掘用户搜索主题相关的感兴趣内容。通过深度挖掘算法分析用户搜索记录、查询历史以及用户感兴趣的相关文档视为用户搜索主题数据的来源,进而挖掘兴趣主题。挖掘模型主要采用向量空间模型,将用户搜索主题模型表示成用户搜索主题向量形式。形成主题和用户兴趣关系网,用户搜索主题向量的构造过程:选择一组用户查询词,并对它们进行深度挖掘分类,最后用它们构造用户搜索主题特征向量,进而分析用户兴趣点。结合用户随着时间的变化,以及过程中有不用的搜索词,以及无关的搜索噪声词去掉,调整兴趣度,用户搜索主题需要具有更新学习机制,动态跟踪了用户兴趣变化趋势。该用户搜索主题研究过程克服了数据稀疏、类别偏差、扩展性差等缺点。实验结果表明,该模型识别用户搜索主题准确率良好。We mainly study that the content of the interest related to the user’s search topic based on deep learning is mined.Through deep mining algorithm,the users’search records,query history and relevant documents that the user is interested in are analyzed as the source of users’search subject data to mine the interest subjects.The mining model mainly uses the vector space model to represent the user search topic model in the form of user search topic vector.The construction process of user search topic vector is as follows:selecting a group of user query words,then mining and classifying them in depth,finally using them to construct user search topic feature vector to analyze user interest points.Combined with the changes of users over time,the different search terms in the process,and the irrelevant search noise words removed,the degree of interest adjusted,users need to have an updated learning mechanism to dynamically track the changing trend of users’interest.The research process of user search topic overcomes the shortcomings of data sparsity,category deviation and poor scalability.The experiment shows that the model has better accuracy in identifying users’searching topics.

关 键 词：深度学习 用户搜索主题 用户模型 挖掘兴趣 个性化搜索 

分 类 号：TP39[自动化与计算机技术—计算机应用技术]