5:15 PM - 6:45 PM
[HDS10-P06] Time Series Analysis of the 2024 Noto Peninsula Earthquake based on Newspaper Articles
-Development of a Corpus of Disaster Words Using Natural Language Processing-
Keywords:local disaster management plan, Noto Peninsula Earthquake, natural language processing, disaster word corpus
A lot of time was spent on emergency and recovery response against the Noto Peninsula Earthquake that occurred on the first day of 2024. It was also reported that there is no plan for road opening. The Basic Law on Disaster Countermeasures insists that prefectural and municipal governments should prepare local disaster management plans, which must be reviewed and revised as necessary. Although all prefectures have prepared regional disaster prevention plans, and 35 organizations have revised their plans 39 times in fiscal year 2008, it is necessary to verify whether these plans will function in the event of a disaster. Therefore, there is a need for research on how to utilize and reflect the actual conditions of past disasters to make local disaster prevention plans more comprehensive and effective. In this study, we conduct a basic analysis as a preliminary step.
We decided to utilize Newspaper articles for understanding the actual disaster situation. Newspaper articles are used because of their immediacy, because they deal with the topic at the time, and because they allow for time-series analysis by days. Previous studies using newspaper articles to analyze the actual status of disasters include an analysis of the Great East Japan Earthquake disaster reports by individual attribute and an analysis of newspaper reports on lifeline damage in the Kumamoto earthquake. In both cases, the articles on disasters were selected as survey targets by extracting articles containing words determined by the analyst, but the analyst's subjectivity was included in the process of extraction. It is also possible that there are omissions of articles that do not contain the word but should be included in the survey as related articles. To ensure reproducibility, we developed a corpus consisted of disaster words to extract newspaper articles.
Since local disaster management plans follow the national disaster management plan, we first sorted compound words and words from the basic disaster prevention plans into parts of speech using natural language processing and developed the resulting corpus into a disaster word corpus. Referring to the Disaster Word Corpus, we extract newspaper articles about the 2024 Noto Peninsula earthquake that contain words from the corpus. Natural language process is conducted on the extracted disaster articles to elucidate the actual situation of temporal changes in the situation, based on changes in the words that appear in the articles. This data processing is used to analyze when and in which scope of work the items listed in the local disaster prevention plan were implemented, and to evaluate the comprehensiveness and feasibility of the plan.
In the future, we will continue to improve the accuracy of the article extraction method using the Disaster Word Corpus and to study the validity of the evaluation method. Furthermore, we are planning to validate the comprehensiveness of local disaster management plan by comparing with the results of the time-series analysis of actual disaster response at 2024 Noto Peninsula earthquake. To analyze the characteristics of the plans for each municipal, the local disaster management plans are also examined by the same method, and we will identify omitted tasks in the plans.
We decided to utilize Newspaper articles for understanding the actual disaster situation. Newspaper articles are used because of their immediacy, because they deal with the topic at the time, and because they allow for time-series analysis by days. Previous studies using newspaper articles to analyze the actual status of disasters include an analysis of the Great East Japan Earthquake disaster reports by individual attribute and an analysis of newspaper reports on lifeline damage in the Kumamoto earthquake. In both cases, the articles on disasters were selected as survey targets by extracting articles containing words determined by the analyst, but the analyst's subjectivity was included in the process of extraction. It is also possible that there are omissions of articles that do not contain the word but should be included in the survey as related articles. To ensure reproducibility, we developed a corpus consisted of disaster words to extract newspaper articles.
Since local disaster management plans follow the national disaster management plan, we first sorted compound words and words from the basic disaster prevention plans into parts of speech using natural language processing and developed the resulting corpus into a disaster word corpus. Referring to the Disaster Word Corpus, we extract newspaper articles about the 2024 Noto Peninsula earthquake that contain words from the corpus. Natural language process is conducted on the extracted disaster articles to elucidate the actual situation of temporal changes in the situation, based on changes in the words that appear in the articles. This data processing is used to analyze when and in which scope of work the items listed in the local disaster prevention plan were implemented, and to evaluate the comprehensiveness and feasibility of the plan.
In the future, we will continue to improve the accuracy of the article extraction method using the Disaster Word Corpus and to study the validity of the evaluation method. Furthermore, we are planning to validate the comprehensiveness of local disaster management plan by comparing with the results of the time-series analysis of actual disaster response at 2024 Noto Peninsula earthquake. To analyze the characteristics of the plans for each municipal, the local disaster management plans are also examined by the same method, and we will identify omitted tasks in the plans.