A Multi-Agent System for Extracting and Querying Financial KPIs and Guidance

View a PDF file from the paper entitled Irorrhoid structuring: a multi -agent system for extracting the main performance and guidance indicators, by Chanyeol Choi and 10 other authors

PDF HTML (experimental) view

a summary:Extracting organized visions and quantity of unorganized financial files is necessary in investment research, however it still takes a long time and the intensity of resources. Traditional methods in practice depend significantly on heavy manual operations, which limits expansion and delay in research progress. In this paper, we suggest an effective and developed method to extract quantitative insights of the undemocratic financial documents, and to take advantage of a multi -agent system consisting of large language models. Our proposed multi-agent system consists of specialized factors: \ Inus {Extraction Agent} and \ Inur {Text to-SQL}. \ Textit {Extraction Agent} determines the main performance indicators of non -structured financial text, determines its formats, and achieves its accuracy. On the other hand, \ Textit {Text-To-SQL} creates SQL phrases that can be implemented from natural language queries, allowing users to accurately access the data organized without the need to know the database scheme. Through experiments, we make it clear that our proposed system effectively transforms the non -structured text into accurate data accurately and allows the main information to be recovered. First, we make it clear that our system achieves a resolution of about 95 % in converting financial deposits into organized data, and usually matching the level of performance by human meal. Second, in the humanitarian evaluation of the task of retrieval – where natural language queries are used to research information from organized data – 91 % of responses were ranked as correct by human residents. In both evaluations, our system is well dependent on the types of financial documents, which provides a constantly reliable performance.