Automatic Classification of Cross-Domain Opinions

Rafael Guzman Cabrera


It is increasingly common for internet users to have access to blogs and social networks. On these sites it is common to issue opinions. Opinions allow people to measure the perception of a particular topic or product. When the number of opinions is very large, its analysis becomes more complicated and it is generally sought to resort to tools that allow this task to be carried out automatically. In the present work the automatic categorization of text opinions is carried out. These opinions correspond to four products: books, DVDs, kitchens and electronics. You have both positive and negative opinions. Categorization results are presented using cross domains as training and testing, using different learning methods and are complemented with similarity graphs, which allow us to have a visual reference of the proximity of the language between the different domains under study. The results obtained allow us to see the feasibility of the proposed methodology.


Opinion classification, machine learning, subjective

Full Text: PDF (Spanish)