APP标签挖掘及应用系统的设计与实现

发布时间：2018-06-13 06:36

本文选题：标签挖掘 + 垂直搜索　；参考：《哈尔滨工业大学》2013年硕士论文

【摘要】：随着智能手机的广泛流行，移动互联网正不断改变着人们的生活方式，将用户从PC端带到了移动端，这一切都是因为手机应用程序(Application)的丰富，使移动终端基本上具有电脑的所有功能，并且使用更加便捷；但是随着App数量的不断增长，用户对App的检索需求也不断加大，传统的搜索引擎很难准确的返回用户所需要的信息，，浪费了大量时间。本文的主要研究内容是通过文本挖掘的方法从网页信息与查询日志两方面获取App的概念体系、属性及语义三种不同的标签，并最终通过这些标签构建一个面向App领域的垂直搜索应用，进行App查询和推荐服务。 App的概念体系标签挖掘是通过获取App网站导航栏信息的方式实现的，通过对导航栏的解析可以获得App的分类信息，之后通过本文提出的一种基于余弦相似度和同义词扩展的分类体系融合策略将所有分类信息进行合并，使得每一个App都有自己的分类体系；对于属性标签，是将抓取到的App简介信息作为文本，通过关键词抽取技术获得而成的；语义标签的获取则是通过查询日志实现的，使用查询日志中的Query、Titile、Url和点击行为等信息，通过本文提出的一系列规则和过滤策略使得查询Query和App一一对应起来，再通过分词、词法分析、句法分析等手段对Query进行处理，获取其语义标签。以上标签挖掘方法不需要人工参与，节省了大量资源开销。整个应用系统采用Django框架进行搭建，并通过Lucene全文搜索工具包对标签数据构建索引，实现了具有语义分析功能的检索系统。经过功能测试，满足人们对App领域检索的需求。
[Abstract]:With the popularity of smartphones, the mobile Internet is changing the way people live, bringing users from the PC to the mobile, all because of the abundance of mobile applications. The mobile terminal has almost all the functions of the computer, and it is more convenient to use. But with the increasing number of App, the search demand for App is also increasing. Traditional search engines are difficult to accurately return the information users need, wasting a lot of time. The main research content of this paper is to obtain three different tags of App from two aspects: Web page information and query log through text mining. Finally, a vertical search application oriented to the App domain is constructed through these tags, and the App query and recommendation services are carried out. The concept system tag mining of App is realized by obtaining the information of the navigation bar of the App website. The classification information of App can be obtained by analyzing the navigation bar, and then the classification information is merged by a classification system fusion strategy based on cosine similarity and synonym extension proposed in this paper. So that each App has its own classification system; for attribute tags, the captured App information is taken as text, obtained by keyword extraction technology; semantic tags are obtained through query logs. Using query Titileurl and click-behavior information in query log, query and App are matched one by a series of rules and filtering strategies proposed in this paper, and then query is processed by participle, lexical analysis, syntactic analysis and so on. Gets its semantic tag. The above label mining method does not need human participation, saving a lot of resources. The whole application system is constructed with Django framework, and the label data is indexed by Lucene Full-text search Toolkit. The retrieval system with semantic analysis function is realized. Through the function test, satisfy the people to the App domain retrieval demand.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TP391.1

【相似文献】