A Reactive Search Re-ranking model based on Images of User Interest

The volume and variety of data on web is increasing day by day, which is making the task of search engines difficult. The search engines typically use the text, inlinks and outlink to rank the web pages based on the user query. The techniques of Search Engine Optimisation like keyword stuffing, link farming etc made the least relevant pages to get a good position when ranked by search engines that puts the user in trouble. The user would have to waste their huge amount of time toggling between pages to find the relevant information. This problem can be surmounted by the use of multiple information modalities like images, videos etc along with the text to judge the relevancy of the web page for a user query. The paper explains how the images in the web pages can be used to re-rank them according to user"s interest. The user would have to provide a user click feedback for each query and the system would re-rank the web pages based on the user click. The method was tested on a set of real time sample queries and a better user satisfaction was achieved.