随着信息的海量增长，互联网环境以及用户个人设备如PC、手机中的数据量急剧增加。作为应对互联网信息过载的重要方式，网络搜索引擎在用户检索信息过程中发挥的作用不可替代，也因此成为互联网的重要入口。同样，面对个人设备中越来越多的文档、图片、视频等，用户越来越依赖桌面搜索引擎来满足信息需求。

桌面环境的特性对桌面搜索引擎提出了新的要求，例如桌面文档频繁变化的特性对桌面索引更新的实时性和系统资源占用提出更高要求。此外，随着移动互联网和手机硬件的迅速发展，用户搜索手机本地信息的需求逐渐凸显，如文档、短信、邮件等的搜索，研究如何搜索手机本地的信息十分必要。

本文针对桌面搜索的若干关键技术展开研究，包括文件系统事件实时监听、桌面索引实时更新、面向桌面搜索的查询词推荐，最后实现了一个更好地适应桌面特性的桌面搜索系统，并进行了将PC平台搜索引擎移植到手机平台的工作。

具体而言，本文的主要研究工作及贡献如下：

（1）     本文研究了Linux内核文件系统变化通知机制，提出了一种Linux文件系统事件实时监听方法。桌面文档频繁发生新建、删除、修改、移动、复制等变化，实时更新桌面索引首先需要实时感知桌面文档的变化，即对文件系统事件进行实时监听。实验表明，本方法监控效率高、能够实时、无遗漏监听文件系统事件。

（2）     基于Linux文件系统事件实时监听方法，本文提出了一种基于文件系统事件监听的索引实时更新方法，重点研究了如何根据不同的事件类型和主体实时、高效地更新桌面索引。实验表明，本方法索引更新速度快、系统资源占用少。

（3）     本文将查询词推荐引入桌面搜索系统，提出了一种面向桌面搜索的查询词推荐方法。

（4）     基于上文提出的方法，设计并实现了一个桌面搜索系统，更好地满足了桌面环境的特性。

With the amount of data increased with massive growth, data in Internet and personal equipment such as PC and cellphone had rapid growth. As an important way to deal with Internet information overload, Internet search engine plays an irreplaceable role in the user retrieve information process so that it becomes an important entrance to the Internet. In the meantime, personal equipment has stored more and more pictures, documents and videos, users have become much more relied on search engine to meet their demands.

The characteristics of the desktop environment put forward some new requirements to desktop search engine. In addition, desktop search demand on the mobile phone increasingly appeared。

 This paper searched a number of key technologies of desktop search engine, including file system events monitoring, desktop index real-time updating based on file system events monitoring, query term recommendation facing desktop search, porting search framework from PC to mobile phone and so on. In the end, this paper implemented a desktop search system based on ElasticSearch search engine.

The main research work and contributions of this paper are as follows:

（1）This paper studied the Linux kernel file system change notification mechanism and proposed a Linux file system event listener method. Experimental results showed that this method had high efficiency and can monitor file system events in real time and without leakage.

（2）This paper presented an index real time updating method. It focused on how to efficiently update the desktop index depending on the event’ types and subjects. Experiments showed that this method updated index faster and used less system resources.

（3）This paper introduced query term recommendation into desktop search system and put forward a desktop search engine query term recommendation method for desktop search.

（4）Based on the above methods, this paper designed and implemented a desktop search system to better meet the characteristics of the desktop environment.

（5）This paper also did the job of porting ElasticSearch search engine to Android platform.