基于位置的移动web搜索方法研究

发布时间：2018-04-18 23:40

本文选题：移动搜索 + 位置　；参考：《华中科技大学》2012年硕士论文

【摘要】：随着移动设备的普及和无线网络的演化，移动搜索逐渐扩大其在搜索市场的份额，也成为搜索研究的一个重要领域。当前的移动用户终端搜索应用或照搬web搜索引擎算法，，或仅仅对位置的进行简单计算与排序，其结果排序并不能准确地反映用户与目标之间的实际到达距离，用户不能快捷的访问与自己距离较近的目标，从而降低用户搜索体验。本文研究的基于位置的移动web搜索方法，就是为了细化搜索，将位置的概念引入到移动搜索中来。对于移动web搜索方法中位置信息的使用和处理，主要有如何获取位置信息以及如何使用位置信息两方面。本文即是从这两方面出发，首先使用CRF（条件随机场）方法提取页面内容中的地名词汇，使用中文地名命名规则来去除地名歧义，使用决策树算法来排除地名干扰，以此确定页面内容描述的位置信息；然后使用一种结合页面重要度和页面位置信息的综合评分机制来对常规web搜索引擎的初步检索结果进行基于位置的重排序，使排序结果的前N项具有更高的相关程度，从而提高用户搜索体验。本文还为两方面的研究分别设计了实验来比较本方法的性能。实验结果表明，结合CRF、命名规则和决策树的页面内容位置抽取方法平均性能较好，结合重要度和位置信息的综合评分机制在精确率上的性能较好。
[Abstract]:With the popularity of mobile devices and the evolution of wireless networks, mobile search has gradually expanded its share in the search market, and has become an important area of search research.The current mobile user terminal search application or copy of web search engine algorithm, or simply calculate and sort the location, the result sort can not accurately reflect the actual arrival distance between the user and the target.Users can not quickly access their own close to the target, thus reducing the user search experience.The mobile web search method based on location, which is studied in this paper, is to refine the search and introduce the concept of location into mobile search.For the use and processing of location information in mobile web search methods, there are two aspects: how to obtain location information and how to use location information.From these two aspects, this paper first uses the CRF (conditional random field) method to extract the toponymic vocabulary from the page content, uses the Chinese toponymic naming rules to remove the geographical names ambiguity, and uses the decision tree algorithm to eliminate the geographical names interference.Then, a comprehensive scoring mechanism combining page importance and page location information is used to reorder the initial retrieval results of conventional web search engine based on location.The first N items of sorting results have higher correlation degree, thus improving the user search experience.Experiments are also designed to compare the performance of this method.The experimental results show that the average performance of the page content location extraction method based on CRF, naming rules and decision tree is better, and the comprehensive scoring mechanism combined with importance and location information has better performance on accuracy rate.
【学位授予单位】：华中科技大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TN929.5

【参考文献】