Identifying the factors of country risk fluctuation from news text data using natural language processing
Sehwan Chung1, Jungyeon Kim1, Seokho Chi1,2, Du Yon Kim3
1 Seoul National University, Korea, Republic of (South Korea)
2 Institute of Construction and Environmental Engineering, Korea, Republic of (South Korea)
3 Kyungil University, Korea, Republic of (South Korea)
DOI: 10.35490/EC3.2023.171
Abstract: Due to the uncontrollability of country-level risk, project participants like contractors should keep monitoring the current situations of the host country. News articles might be good sources to extract issues preceding country risk fluctuation. This study proposes a framework to recognize risk-related issues from news content. S-BERT vectorizes news texts, and DBSCAN groups them into several topics. Topics from different timespans are compared to distinguish risky issues. Preliminary results show the proposed framework can extract specific topics from news articles. It is expected that participants of international projects utilize the framework to recognize the situations of the host country.
Keywords: DBSCAN, Host Country Risk, International Construction, natural language processing, News Text, Risk Fluctuation Factors, S-BERT, Topic Extraction
