Automatically measuring the quality of user generated content in forums

Abstract

The amount of user generated content on the Web is growing and identifying high quality content in a timely manner has become a problem. Many forums rely on its users to manually rate content quality but this often results in gathering insuffcient rating. Automated quality assessment models have largely evaluated linguistic features but these techniques are less adaptive for the diverse writing styles and terminologies used by dierent forum communities. Therefore, we propose a novel model that evaluates content, usage, reputation, temporal and structural features of user generated content to address these limitations. We employed a rule learner, a fuzzy classier and Support Vector Machines to validate our model on three operational forums. Our model outperformed the existing models in our experiments and we veried that our performance improvements were statistically signicant.