AN EFFICIENT IMAGE BASED DOCUMENT RETRIEVAL USING SEMANTIC ATTRIBUTES
Abstract
Search engines and their ranking algorithm perform an important role in World Wide Web. Each Search engine has its own ranking algorithm. The ranking algorithm is used to rank the web pages and display it to users. But, user expectation is highly exceed than search engine experience. Traditional web search engine mainly uses text and in- link or out-link in the web pages to do the ranking of the web pages. But, they do not consider the multimedia in the web pages to find relevant documents for a given query. Instead of this approach, a measure of agreement is estimated between the keyword provided by user and text portion of each web page. In this work, we propose an image based re-ranking system and also semantic attribute. The re-ranking system uses a pure text based search engine to find a primary set of candidate documents for a given query. Then, the candidate set is reranked using images contained in the web pages. The semantic attributes are used to focus the common features of real world image to improve object classification performance. The results of this experiment confirm the improvement in accuracy produced by our image-based re-ranker over a pure text-based system.
Author
S.Porpetchi
Dr.S.V.Manisekaran
Download
