PSEWEIP : Personalised Search Engine and Web Experience Improvement Project

Abstract

The fact that the size of the World Wide Web is huge is common trivia amongst all of its users. For this reason, the research presented in this project is aimed at improving the current users browsing experience. The main improvement we want to provide to the current user's browsing experience includes giving them the possibility of having search engine results ordered by relevance to their browsing context. On the other hand, we also want to provide other additional features that can also yield to an improvement in the browsing experience namely the possibility of having access to the queries used to visit a particular document, and the possibility of storing notes for further future access on documents which the user visits. In order to be able to provide context based sorting to users, we modelled phrases found in documents that were visited since these can be potential context describers. Afterwards, we allow the user to pick the document that best fits his context, such that a sorting of the search engine results based on phrase based similarity measures is performed. Phrases are used in the similarity measures since they are seen as better describers of context. With regard to the context based presentation of results, the results obtained are promising as we have been capable of providing the users with relevant results that originally had very low ranking in the search engine's result set. Furthermore, we have also increased the rate of relevant documents in the top results, though some irrelevant results were still returned, so there is still a need of further improvement. Nonetheless the users' feedback was very positive about this feature and showed that it would be an asset to find more relevant results without reformulating a query multiple times. The other features presented in this project also received a good feedback from the users, who have shown interest in knowing which query was used to access which document and also an interest in being able to store notes on particular documents for future referral. I am hopeful, therefore, that the study presented in this research has provided a further contribution to the areas of context based document sorting using phrase based similarity measures, phrase based similarity, and web information retrieval.