Assistente financeiro virtual baseado em processador de linguagem natural, reconhecimento de voz e XBRL

• Ano de defesa: 2016
• Autor(a) principal: ABRAÃO, Adalberto Alves
• Orientador(a): SILVA, Paulo Caetano da
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Salvador
• Programa de Pós-Graduação: Sistemas e Computação
• Departamento: Sistemas e Computação
• País: Brasil
• Palavras-chave em Português: Assistência financeira virtual; XBRL; Taxonomias financeiras; Arquitetura para assistentes virtuais de domínio restrito; Assistente de propósito especifico; Linguagem natural
• Área do conhecimento CNPq: Ciências Exatas e da Terra
• Link de acesso: http://teste.tede.unifacs.br:8080/tede/handle/tede/573

When an investor want to invest in a company, it is necessary to know the current and past financial health of this company. Extracting this financial information from an information system that uses recognized taxonomy by the financial markets, using natural language, is a facility for the investor who has no knowledge and no expertise in computer science. As alternative to assist people in a particular area of knowledge without requiring them prior knowledge in information technology, there are the virtual assistance software. However, it was observed that there is not available a software of virtual assistance that answers made user questions in natural language about financial health of companies based on available information in electronic reports. This paper presents a virtual assistance tool to maintain a dialogue with the user using natural language, in English or Portuguese, to answer financial questions about the companies that acts in the US financial market. This tool, named Financial Virtual Assistant, recognizes the user questions through natural language processor and provides responses based on available information in electronic financial statements in XBRL technology, in accordance with US-GAAP taxonomy. The Assistant implementation, its architecture, based on services, the selection and configuration of the used natural language processor service, are also presented. As a result of the tests and evaluations of the assistant implementation, it was observed that the implemented Financial Virtual Assistant acted as proposed, and correctly answered user questions, and presented compatible performance results with the available virtual assistant software of the general purpose. This work contributes to the field of virtual assistance software and information extraction area of electronic forms, with an architecture for the closed domain assistants, based in services and natural language processing.