自然言語処理の応用による持続可能な社会シナリオの構造化支援

抄録

Many scenarios have been developed by organizations to support their strategic planning toward achieving sustainable futures. While scenarios are typically presented in text format, the authors have developed the scenario structuring method to visualize the logical structure of a scenario in graph format for assisting mutual understanding and communications among stakeholders. However, the most critical problem is that executing this method is very timeconsuming because scenario structuring requires in-depth reading by humans. To tackle this problem, in this paper, we propose a method for supporting scenario structuring using natural language processing. We used the state-of-the-art pretrained language models (i.e., ALBERT, ERNIE, and DeBERTa) to semi-automate scenario structuring tasks. Results showed that the proposed classifier system reached an F1 score of 87.5 in finding two sentences that are semantically related to each other in a test scenario.