Evidências de qualidade de atributos textuais na web 2.0

Abstract

The advent of the Web 2.0 changed the way users interact with web applications on the Internet. Nowadays, such users are not only passive content consumers but also content creators on the Web. The result of this collaborative form of content creation, known as social media, is one of the main factors behind the massive popularity (in amount of users and data) of the Web 2.0. One shortcoming of social media is the possible lack of quality of the content generated by users. In fact, one recent study pointed this lack of quality as one of the reasons why Information Retrieval (IR) services do not effectively explore social media yet. In order to provide insights on the matter, this dissertation presents a comparative study of the quality of different textual features on the Web 2.0. A textual feature is a region of a web page with a well defined topic and functionality. We studied the textual features Title, Tags, Descriptions and Comments on LastFM, CiteULike, Youtube and YahooVideo. Our study consist of three different comparisons. Firstly, we compare the quality of features regarding three aspects of quality, namely: usage, amount of content, syntactic correctness, descriptive quality and discriminative quality. After our characterization, we compare textual features when applied to the IR tasks of object classification and recommendation. Lastly, a study with 17 volunteers was performed in order to compare how users perceive the quality of features. Our work extends previous work which focus mostly on Tags. The results presented on this dissertation can be explored by Web 2.0 providers and designers in order enhance their IR services and develop Web 2.0 applications considering the benefits and shortcomings of each textual feature.