| dc.description.abstract | In a daily basis, people face judgments or ideas regarding certain issues in different means of communication, such as, news editorials, magazines, forum threads, e-mails, blogs, etc. In this context, these opinions begin to strongly influence people s decisions and can often be useful to find out what people think about a particular topic. Then, how to keep track of the mass opinions on the web? Without a thorough assessment of the clues presented in a text this task becomes impossible. Given the importance of these opinions in the generation of ideas, feelings and even impressions, this paper work focuses on the study of lexical patterns in opinion texts, based on the theories presented by Wiebe e Mihalcea (2006), Kim and Hovy (2006), Wilson et al (2005), among others. The corpus used in this study and the method that identifies expressions that denote evaluation have been taken from Folha de São Paulo, a newspaper database. After the delimitation of the expressions and the use of computational tools, two evaluative glossaries have been defined, a positive and a negative one. Thus, this study aims at contributing with the identification of opinions in other texts/contexts and at providing a basis for future work in the field of sentiment analysis. | |
