Tracking Context in Conversational Search: From Utterances to Neural Embeddings

• Autor(a) principal: Ferreira, Rafael André Henriques
• Data de Publicação: 2020
• Tipo de documento: Dissertação
• Idioma: eng
• Título da fonte: Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
• Texto Completo: http://hdl.handle.net/10362/116770
• topic: Conversational Search; Multi-turn Question Answering; Conversational Context; Information Retrieval; Query Rewriting; Ranking; Domínio/Área Científica::Engenharia e Tecnologia::Engenharia Eletrotécnica, Eletrónica e Informática

The use of conversational assistants is becoming increasingly more popular among the general public, pushing the research towards more advanced and sophisticated techniques. Hence, there are currently a number of research opportunities to extend the comprehension and applicability of these tasks in everyday systems. These conversational assistants are capable of performing various tasks, such as chitchatting, internal device functions (e.g., setting up an alarm), and searching for information. In the last few years, the interest in conversational search is increasing, not only because of the generalization of conversational assistants but also because conversational search is a step forward in allowing a more natural interaction with the system. To build a system such as this, many components need to work together, since in a conversation, the importance of context is paramount to retrieve the best answers to the user’s questions. In this thesis, the focus was on developing a conversational search system that aims to help people search for information in a natural way. In particular, this system must be able to understand the context where the question is posed, tracking the current state of the conversation and detecting mentions to previous questions and answers. We achieve this by using a context-tracking component based on neural query-rewriting models. Another crucial aspect of the system is to provide the most relevant answers given the question and the conversational history. To achieve this objective, we used state-of-the-art retrieval and re-ranking methods and expanded their architecture to use the conversational context. The results obtained with the system developed achieved state-of-the-art when compared to the baselines present in TREC Conversational Assistance Track (CAsT) 2019.