Conversational Aware Suggestion System

• Autor(a) principal: Ferreira, Diogo Filipe Peixoto
• Data de Publicação: 2019
• Tipo de documento: Dissertação
• Idioma: eng
• Título da fonte: Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
• Texto Completo: http://hdl.handle.net/10773/28331
• topic: Pervasive computing; Context-awareness; Information retrieval; Suggestions system; Text mining; Reinforcement learning

Over the last few years, pervasive systems have experienced some interesting development. Nevertheless, human-human interaction can also take advantage of those systems by using their ability to perceive the surrounding environment. In this dissertation, we have developed a pervasive system - named ConversationaL Aware Suggestion SYstem (CLASSY) - which is aware of the conversational context and suggests the users potentially useful documents or that, somehow, save time executing a specific task. We have also proposed two different approaches - the Neighborhood one, that uses semantic similarity, based on proximity data in order to classify the relationship between tokens; and the Reinforcement Learning one, that uses implicit feedback associated with each suggestion as a source of knowledge that can be used to improve the system's performance over time. The conducted tests showed that these two approaches not only enhanced the pervasive behavior of the system, but also increased its global performance. A case study regarding the importance of feedback on context-limited environments was also carried out, whose results showed that it is still a useful source of knowledge regardless the conversational environment's characteristics.