Recomendação automática de atrações turísticas a partir da análise semântica de comentários de usuários de plataformas colaborativas: uma aplicação da Framenet Brasil

Abstract

The present work is part of the FrameNet Brasil project (FN-Br) (SALOMÃO et al., 2009) and, therefore, fits in the theoretical-methodological discussions related to Frame Semantics (FILLMORE, 1982) and Qualia relations (PUSTEJOVSKY, 1995), since it proposes the applicability of computational resources, from the modeling of domainspecific frames. In this context, this thesis has as its basis of investigation two fundamental questions: (i) how can the tourist make the planning of a trip more effectively, without having to search in several travel guides, thus demanding less time; (ii) how tourists can get more specific information about tourist attractions, such as “attraction for children” or “restaurant with accessibility for the elderly”. Thus, these questions guide this work, as they enable the analysis of how Frame Semantics can assist this tourist in the search for a better tourist experience. From this perspective, this thesis aims to survey frames related to the specific domain of tourism, focusing on the investigation of frames of secondary interest, that is, those that are not restricted to this specific domain, thus, these frames are modeled on the FrameNet Brasil database. This secondary frame survey is taken through the semantic analysis of tourist comments from Google Places. In addition, this thesis proposes the modeling of these frames as well as the modeling of ternary Qualia relations, in order to test the performance of the recommendation system developed by FrameNet Brasil, based on the interaction between the tourist and the chatbot. Therefore, this thesis seeks to demonstrate how frames of secondary interest, from Frame Semantics, together with ternary Qualia relations can assist in the development of computational resources and, more specifically, and provide the tourist with a more effective interaction with the application.