Master Thesis Cross-Media Meta Search: Query Relaxation and Information Integration for Heterogeneous-Media Search Engines

In recent years, the quantity of multimedia contents on WWW has been increasing with improvement in the speed of transmission speed, and the spread of digital cameras. Since the contents are distributed on a lot of websites and expressed by various media, it is important to construct a function which searches the contents effectively. Conventional meta search engines retrieve the results from several search engines and automatically classify them. The meta search engines dispatch queries to same type search engines, and retrieve mono-type contents. Moreover, only the link to a Web page is shown in ranking form as a unified reference result. However, in such conventional meta search engines, the information on various media is simultaneously uncollectible. If a user follows the link of a reference result and peruses two or more Web pages, he can acquire effective information at last. In this paper, in order to solve this problem, we propose a system which dispatch queries to adequate search engines based on the media type. Moreover, we propose the system which extracts only the information in relation to the reference keyword group from the Web page obtained as a reference result, and unifies these automatically. The information over various media relevant to the reference keywords are searched and unified. It can be called output like an encyclopedia using Web page. Users can only input reference keywords and can peruse easily various information currently distributed on Web about the keyword. Users can effort of perusing the Web page of a link place sequentially from the result of higher rank and discovering effective information from Web pages. And also users can peruse the information over reference keywords now at a glance effectively so that it may say that this language is such meanings. In this system, the case where are the reference question by two or more ii keywords is inputted by the user is considered. The search engine used here corresponds to various media. That is, it is the search engine with which the types of a text search, a image search, a music search, etc. We focus on a text search and a image search. we propose the method of transforming a reference question and collecting Web pages efficiently. A text search has many hits of a solution, but it is difficult to acquire effective information. A image search has high precision and can obtain the picture of a …