Uma abordagem multi-visão para a estimativa automática da qualidade de conteúdo colaborativo na Web 2.0

Abstract

The Web contains a new type of repository for the human knowledge where users are able not only to consume, but also to produce content in a much faster and easier manner. However, such freedom also carries concerns about the quality of this content. In this thesis, we propose an automatic quality approach to assess the quality of collaborative generated content. To accomplish this, we adopt a multi-view approach to assess the quality of content, in other words, we apply machine learning (ML) techniques to combine independent assessments regarding di erent sets of semantically related quality indicators (i.e., views) into a single quality value. Then, we perform a thorough analysis of our approach in two di erent domains (Questions and Answer Forums and Collaborative Encyclopedias), which allowed us to better understand when and how the proposed multi-view approach is supposed to improve quality assessment. We also study the impact of the views and the features that compose them in each domain. To summarize, our main contributions include: (1) the proposal of a general multi-view approach that takes advantage of groups (i.e., views) of quality indicators;(2) the proposal of new features in Q&A Forum domain; (3) the application of this approach in 2 domains where we could achieve an improvement of up to 30% in quality assessment over the best baselines methods found in the literature; (4) a throughout feature and view analysis regarding impact, informativeness and correlatedness, considering both domains.