| dc.description.abstract | The rapid expansion of distance education in the country has been the focus of academic research, especially regarding the creation of strategies to avoid high dropout rates for students in distance education courses and disciplines. Within this context, this research is part of a project entitled MAS-EAD (Automatic Mapping of Feelings in Distance Learning). Since 2012, the aforementioned project has investigated strategies for the linguistic-computational representation of the Brazilian Portuguese emotion and opinion lexicon that works as a component of a computational system dedicated to the mining of opinions registered in a virtual teaching and learning environment. Thus, this research aims at describing emotion metaphors in the context of distance learning and, mainly, at including them into the software under development in the MAS-EAD project. The results of this thesis, therefore, contribute to the development and improvement of the sentiment analysis software to be integrated into virtual teaching and learning environments, such as Canvas. The hypothesis is that a research that seeks to enhance artificial intelligence software might have more benefits, in terms of the quantity and variety of linguistic pattern occurrences, by adopting a linguistic approach focused on speech. discursive approach. Therefore, with the background of Corpus Linguistics (BERBER SARDINHA, 2004) and thinking about the linguistic dimension, which is the main dimension addressed in this work, Cameron's Systematic Metaphor is evoked (CAMERON, 2003; BERBER SARDINHA, 2007; CAMERON; DEIGNAN; 2006, 2009; CORTEZ, 2012) to study the aforementioned linguistic phenomenon and describe it. The present research has an interpretive character and adopts a qualitative approach that is anchored in the general methodology proposed in the context of the MAS-EAD project; hence adopting the task-driven research ideology. As an interdisciplinary investigation, the methodology covers three mutually complementary domains. The three domains are typical of research in Natural Language Processing (PLN), as proposed by Dias-da-Silva (1996; 1998; 2003): (i) linguistic, (ii) linguistic-computational; and (iii) computational. In the first domain, theories are studied with a view to the creation of linguistic rules for the domain (ii). Domain (iii) concerns the implementation of these rules in the software, a task attributed to the group of computer experts. The results show that the discursive topics of the recurring tweets refer not only to distance learning and teaching but also to its participants - teachers, tutors -, grades, tests, and activities. Furthermore, in the scenario designed for the small sample represented within the research corpus, the majority of tweets show dissatisfaction of students and former students of the modality regarding these various topics mentioned above. | en |
