: Financial Virtual Assistant

Abstract

— Getting information about the current and past financial situation of a company is important before investing in this company. Extracting this information from an information system that uses taxonomy recognized by the financial markets using natural language, is a facility for the investor who has no knowledge and no expertise in computer science. The virtual assistance software are alternatives to help people in an area of knowledge. However, it is observed that there is no information system that interacts with the user through natural language to answer questions about financial information of companies, based on data available in electronic financial disclosures. This paper presents a computational system of virtual assistance, named Financial Virtual Assistant (FVA), which recognizes user questions relating to financial situation of companies, through text, or voice, in natural language. This system provides answers based on available information in electronic financial statements represented in eXtensible Business Reporting Language (XBRL) technology. The system's implementation is based on a proposed architecture for specific purpose virtual assistants that uses a Natural Language Processor (NLP) and the domain information services. Details of this architecture, the system’s implementation and the used natural language processing configuration are presented. During testing, the Assistant correctly answered all financial questions about certain companies in a compatible average period with available generic virtual assistants in the market. Besides that, the Assistant could to talk to the user, simulating a conversation between people.