Computational Science Center Seminar

In recent years, the web has become an essential resource for obtaining information associated with any topic or domain. The amount of texts produced by interactions on social media, blogs, url's, etc., has made essential the use of advanced techniques to be able to understand and obtain valuable information from these large volumes of data. Considering the complexity and richness of the web information, the use of graph-based representations is a growing area of study which aims is to discover novel and insightful knowledge from data. The use of this kind of representations on text documents has a wide range of applications like social science, homeland security, finance, healthcare, web analysis, text linguistics, etc. In this presentation, there are discussed two graph-based approaches used to discover the author and demographics aspects of text documents in the context of the authorship identification and author profiling tasks. The main focus of both approaches is the extraction of relevant linguistic features based on network analysis techniques to create novel supervised learning models that can tackle the analysis of texts in an intuitive way, taking into account the syntactic structure of text documents (the way texts were written).