Research On The Retrieval Of Location-related Web Search

Confronted with the huge scale of online information, it has been a critical problem how to find desirable information effectively and timely. Recently, Web Information Retrieval (IR) which is expected to solve the matter has become the focus of many researches. The retrieval results should be relevant not only to the page content but also to the location from users in more and more applications. Location-related Web search can improve the precision of search results and has wide application.According to query interfaces and application requirements, there are two typical kinds of applications for location-related Web search. One is based on text and the other is based on image. Text-based retrieval is finding objects located in or near some location, in which location information is included and can be described easily. Image-based retrieval mainly is finding where/what is the object. If the user cannot describe the location in simple words or doesn't know how to describe it, he can take photos of the location and send to our server. Then the server will return relevant pages/information to the user after matching images in image databases.This thesis mainly focuses on the above two kinds of location-related Web search. For the difficulties among them, we propose the corresponding solutions, including correspoinding index structures and the algorithm of image matching. And we present the systemic solutions and build up the prototypes to validate them.The work and contributions in this thesis include:1. A kind of hybrid index structure is proposed to solve the existing problems of previous text-based solutions. Previous solutions treat the input keywords as common textual keywords, and retrieve pages via keyword...