[2Win5] Poster session 2

[2Win5-82] Attempt to Extract and Analyse Causality from Agricultural Newspapers

〇Tetsuo Katsuragi³, HIROKI SAKAJI¹, Akio Kobayashi³, Shotaro Mori⁴, Masahiro Otomo³, Junichi Ishihara³, Masahiro Suzuki², Takahiro Kawamura³, Itsuki Noda¹ (1.Hokkaido University, 2.Univ. of Tokyo, 3.National Agriculture and Food Research Organization, 4.HIMIKA, Inc.)

Keywords:Causality, Text Mining

In the Cabinet Office's Project of ``programs for Bridging the gap between R & d and the IDeal society (society 5.0) and Generating Economic and social value,'' the AI Agricultural Society Implementation Project aims to establish AI technologies to compensate for the decrease in agricultural labor force caused by the declining number of farmers.
Among the challenges, the reduction of prefectural extension workers, who directly instruct farmers, is significant.
This project is working on creating AI systems capable of answering questions like an extension agent, utilizing Large Language Models (LLM).
At the National Agriculture and Food Research Organization (NARO), we are working to develop a system that can effectively respond to queries related to farming management—a key support need for extension agents.
Our approach involves constructing datasets that represent the impact of weather on market conditions and aligning these data with their corresponding natural language expressions.
Previous research has highlighted a critical shortage of such natural language data. In this study, we expand the existing dataset and investigate the correspondence between natural language expressions and actual market data, thereby establishing a foundational dataset for the future development of a Q
&A system to support farming management guidance.