Presentation Information
[4Yin-B-01]LLM-based Automatic Knowledge Graph Construction and Integration for Question Answering on Securities Reports
Yuya Fujiwara1, 〇Hiroshi Sasaki1 (1. The Japan Research Institute, Limited)
Keywords:
Large Language Models,Knowledge Graphs,Annual securities reports
Annual securities reports (ASR) are important documents for understanding a company's business status and financial statements. However, these reports are often too complex and verbose for individuals to fully comprehend and utilize. To address this issue, this paper proposes a novel method that utilizes Large Language Models (LLMs) and knowledge graphs to facilitate the use of ASR. The proposed method involves two main functions: an LLM-driven knowledge graph construction to extract information from ASR, and an autonomous question-answering for ASR empowered by the constructed knowledge graphs. We verified the effectiveness of our proposed method through a quantitative evaluation using a competition dataset for a question-answering task on tabular data within ASR.