Chatbot Automóvel Adaptável

• Autor(a) principal: Tiago Miguel Moreira Ferreira
• Data de Publicação: 2017
• Tipo de documento: Dissertação
• Idioma: eng
• Título da fonte: Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
• Texto Completo: https://hdl.handle.net/10216/106510
• topic: Engenharia electrotécnica, electrónica e informática; Electrical engineering, Electronic engineering, Information engineering

With the increasing number of everyday objects having an Internet connection and being able to send and receive data, the so called IoT (Internet of Things) -- the need to have systems that interact with such objects has also grown. This need has grown so much that at this time most manufacturers already ship software designed to interact with their products. However, this typically requires the user to install a separate application for each product and, most of the times, to navigate through complicated menus just to perform some simple tasks. The advent of chat bots has proven to be a worthy topic of research and development, with general acceptance from the target audiences, who view them as a replacement to standard mobile applications, one of the reasons being that bots can be used in their favorite messaging application. Most of the bots available in the market respond only to specific user input; however, with the recent advances in Artificial Intelligence and Natural Language Processing, this approach is being shifted towards engaging conversations with the users, serving as a way to engage the user with the product. This has been proven with the creation of personal assistants by some big companies like Google or Apple. This dissertation aims to create a chat bot that extends the relationship of the users with their cars, allowing a user to easily make location-aware operations, check the car current status and perform operations on the car just by talking with the bot. For instance, the user can locate where the car is parked or even get notified when there is a problem with the mechanical system of the car. This is done by sending the car data through an OBD-II dongle with permanent 4G connectivity to an application server. The server receives user input that is processed using state of the art natural language processing systems as a service such as Amazon Lex, and sends commands back to the car if necessary. The bot learns user patterns by modeling past trips using the ARIMA statistical model. This being a bot that must be able to be used while driving using voice to communicate, raises some challenges mainly on how to interact and direct the conversational flow, understanding the differences between texting and talking, along with accurately and at the right time make proactive suggestions to the user. To evaluate this work, the application was tested with usability tests both internally and externally, where at the end, the participants were asked to fill a questionnaire regarding their experience. The pattern recognition and predictions were also tested against real data.