Department of Computer Engineering
S E M I N A R
Software Design, Implementation, Application, and Refinement of a Bayesian Approach for the Assessment of Content and User Qualities
Computer Engineering Department
The internet is an important resource for obtaining information in every type of fields. After the invention of the second world wide web generation Web 2.0, common people have opportunities to supply contents for the web. This situation causes massive amount of information to be in the web, and makes finding the relevant and up-to-date contents difficult. In order to solve this problem, some filtering mechanisms such as recommender systems based on collaborative filtering, and trust based systems are introduced. The systems consider the past preferences of the users in order to decide the contents to suggest them. The researchers also propose to define some quality levels for the contents in the web. By combining both of the solutions, a novel model based on the user feedbacks is introduced. According to that model, the quality of an evaluator is an important factor which determines the quality of a content in the web. The model uses a Bayesian approach which allows the co-evaluation of an evaluator and a content simultaneously. The Bayesian approach is also suitable for applying the change in qualities over time. In this thesis, we implemented a configurable software which uses that methodology in order to assess the qualities of users and contents. We perform some experiments on movie data set in order to observe the performance of Bayesian co-evaluation according to the baseline approach. The experimental results show that Bayesian co-evaluation increases the retrieval performance based on the baseline approach by 18%. The consistency of the methodology is also tested by applying more experiments with different parameters. The results indicates that our methodology succeeds in identifying the high quality users.
Keywords: Information quality, web 2.0, collaborative systems, recommender systems, collaborative filtering, Bayesian, co-evaluation, user trust.
DATE: 15 August, 2011, Monday @ 14:40