Application of machine learning techniques to the re-ranking of search results

Even though search engines cover billions of pages and perform quite well, it is still difficult to find the right information from the returned results. In this paper we present a system that allows a user to re-rank the results locally by augmenting a query with positive example pages. Since it is not always easy to come up with many example pages, our system aims to work with only a couple of positive training examples and without any negative ones. Our approach creates artificial (virtual) negative examples based upon the returned pages and the example pages before the training commences. The list of results is then re-ordered according to the outcome from the machine learner. We have further shown that our system performs sufficiently well even if the example pages belong to a slightly different (but related) domain.