Analysis of Texts of Fire Accidents in University Chemistry Experiments Using AI Text Mining

Artificial Intelligence (AI) text mining analysis with a "user local" program was conducted on "chemical experiment accident case in 2009". A Japanese document was utilized that summarized the data of accident cases of university chemistry laboratories accompanied by past fires. This document was analyzed using an AI text mining program to check the frequency of word occurrence, relationships between words, etc. Based on the results of this analysis, as well as the opinions of university students concerning these results, we attempted to answer the following research questions. What major factors can be identified about chemical experimental accidents, such as "cause-effect" and "substance-condition"? Do differences exist between statistics and human senses regarding performing chemical experiments? Comparison of the results between text mining and human sense demonstrated that it remains difficult to analyze conditions of specific accidents only from AI text mining results, at least with the aid of present knowledge of chemistry.