基于用户偏好分析的个性化信息检索关键技术研究

发布时间：2018-02-13 16:53

本文关键词： 个性化信息检索用户偏好鼠标滑动满意度意图边界检测社交网络　出处：《苏州大学》2016年硕士论文　论文类型：学位论文

【摘要】：个性化信息检索旨在根据用户不同的信息需求,结合主观价值倾向,给具有不同信息需求的用户返回符合其期望的检索结果。其核心内容在于分析不同用户的个性化特征,结合其自身价值倾向差异,建立个性化检索模型,从而更好的提升用户的检索体验。本文针对个性化信息检索中用户偏好分析任务开展了深入研究,文章的主要内容归纳为以下三个方面:基于“滑鼠行为”量化的检索满意度研究从用户信息检索行为特征入手,提出一种基于“滑鼠行为”能量消耗的满意度评价及量化模型。其中,“滑鼠行为”指用户在网页中的鼠标滑动轨迹以及点击等行为序列。首先,从用户端对信息检索质量评价的主观因素出发,引入“检索满意度”作为衡量信息检索质量的新特征;其次,通过计算用户“滑鼠”过程中的“能量消耗”,分析检索行为特征与用户满意度之间的关联性;最后,基于量化数据建立滑鼠的“能量消耗”对比模型,用于表征用户检索满意度的强弱。这一研究重点关注用户检索满意度的量化分析方法,为信息检索系统质量评价提供了一种新的特征。基于边界检测的查询意图切分技术研究通过分析用户连续查询日志,提出一种查询意图边界识别方法。用户完整查询意图中包含大量的检索习惯及个性化特征,识别查询意图发生改变或转移的临界位置,能够为从完整意图分析层面提取用户的个性化检索偏好提供有效的方法。首先,从大规模样本集中抽取并人工标注部分完整意图片段;其次,结合文本特征抽取方法构建自动化学习模型,包括分类及序列标注的方法,分别使用SVM和CRF模型;最后,通过开发及测试数据集优化学习算法,获得最优的意图切分性能。这一研究借助用户查询日志,通过自动化的意图片段切分技术,为抽取用户的个性化偏好特征提供研究基础和技术手段。用户社交媒体活动与信息检索关联性分析研究通过获取用户在一段时间内的网络活动内容,分析用户在不同时间段内进行信息检索与社交网络活动之间的关联性,建立基于个人文化背景及兴趣倾向的个性化检索模型,改善检索体验。首先,设计实验平台抓取用户在一段时间内的网络活动信息;其次,抽取其中社交网络内容及信息检索相关内容;最后,构建关联性分析模型,度量社交活动与检索行为之间的关联关系,以便利用社交网络信息构建个性化检索模型。这一工作侧重研究用户的文化背景及价值倾向对信息检索行为的影响,以便为利用社交网络信息构建个性化检索模型提供理论依据及研究基础。综上所述,本文针对个性化信息检索中用户偏好分析的任务,从三个方面开展相关工作,旨在为个性化信息检索中的用户偏好分析提供更加新颖、有效的研究思路及技术方法。
[Abstract]:The purpose of personalized information retrieval is to return the desired retrieval results to users with different information needs according to their different information needs and subjective value tendency. The core content of personalized information retrieval is to analyze the personalized characteristics of different users. According to the difference of their own value tendency, a personalized retrieval model is established so as to improve the retrieval experience of users. This paper has carried out a deep research on the task of user preference analysis in personalized information retrieval. The main contents of this paper are summarized as follows: the research of retrieval satisfaction based on "mouse behavior" quantification begins with the characteristics of user information retrieval behavior. This paper presents a satisfaction evaluation and quantification model based on "mouse behavior" energy consumption. Among them, "mouse behavior" refers to the user's mouse track and click behavior sequence in the web page. Starting from the subjective factors of the evaluation of information retrieval quality at the user side, this paper introduces "search satisfaction" as a new feature to measure the quality of information retrieval. By calculating the "energy consumption" of the user "mouse", the correlation between the retrieval behavior characteristics and the user satisfaction degree is analyzed. Finally, a comparative model of the "energy consumption" of the mouse is established based on the quantitative data. This research focuses on the quantitative analysis method of user search satisfaction. It provides a new feature for the quality evaluation of information retrieval system. The research on segmentation of query intention based on boundary detection is based on the analysis of users' continuous query logs. A method for identifying the boundary of query intention is proposed. The complete query intention contains a large number of retrieval habits and personalized features to identify the critical position where the query intention has changed or shifted. It can provide an effective method for extracting users' personalized retrieval preferences from the level of complete intention analysis. Firstly, part of the complete intention fragments are extracted from large scale sample sets and some complete intention fragments are annotated manually. Combining the text feature extraction method to construct the automated learning model, including classification and sequence tagging, using SVM and CRF model respectively. Finally, the learning algorithm is optimized by developing and testing the data set. To obtain optimal intent segmentation performance. This study uses automated intent fragment segmentation technology with the help of user query logs. In order to extract the personalized preference features of users, the research provides the research foundation and technical means. The relevance analysis of user social media activities and information retrieval through the acquisition of users in a period of time network activity content, This paper analyzes the relationship between users' information retrieval and social network activities in different time periods, establishes a personalized retrieval model based on personal cultural background and interest tendency, and improves the retrieval experience. The experiment platform is designed to capture the user's network activity information for a period of time. Secondly, the social network content and information retrieval are extracted. Finally, the relevance analysis model is constructed. The relationship between social activities and retrieval behaviors is measured in order to construct a personalized retrieval model using social network information. This work focuses on the impact of users' cultural background and value orientation on information retrieval behavior. In order to provide theoretical basis and research basis for constructing personalized retrieval model by using social network information. In summary, this paper, aiming at the task of user preference analysis in personalized information retrieval, carries out related work from three aspects. The purpose is to provide more novel and effective research ideas and technical methods for user preference analysis in personalized information retrieval.
【学位授予单位】：苏州大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP391.3

【相似文献】