Web-based profile builder

Abstract

The World Wide Web evolves on a daily basis - today, it is estimated that there are in excess of 8 billion web pages. This massive proliferation of information is distributed across a large number of sources. Seeking the required information and aggregating disparate, yet similar sources can be very challenging. A particular class of information that is heavily sought on the Internet are reviews. Through the assistance of contemporary technologies, the identification of information via public reviews to gain increased insight on services can be considered a relatively modest task. Nonetheless, due to the sheer amount of information available, it is still considered somewhat challenging to establish trustworthy review sources. Another challenge manifests itself in the potential bias towards or against specific services some review sites may exhibit. This disse1iation researches and subsequently proposes a software architecture to improve the typical information gathering experience in the context of service reviews. Since reviews are highly sought after, this is accomplished by harvesting various topic specific review sources in the shape of websites. The use of classification technology is exploited to classify the information retrieved from the reviews and use these results to create an augmented review weighting system. Restaurant reviews were chosen to be used as a proof of concept, on the basis that it is believed that there are still ample opportunities for further exploration in this particular context. End users are allowed to describe the subject matter of interest, and a profile is constructed alongside the display of random reviews and estimated ratings. Through quantitative and qualitative analysis, the prototype's architecture is critically evaluated and its effectiveness measured against typical review systems. The primary results from this analysis indicate that the proposed architecture is capable of handling a substantial amount of information and present highly relevant ratings.