Text mining analysis of teachers' reports on student suicide in South Korea.

A teacher as a suicide prevention gatekeeper has an important role in identifying suicide risks and warning signs in students. After a student's suicide, teachers in Korea have to write a student suicide case report based on their direct and indirect observations. In particular, the section 'characteristic of student suicide' of this report contains valuable information about the suicide; however, it is unstructured, and thus cannot be analyzed using conventional statistical methods. We aimed to identify the characteristics of observed Korean students, who have committed suicide, using text mining techniques as well as to improve our understanding of suicidal behaviors in the school contexts. Therefore, a series of text mining techniques: topic analysis, word correlation, and word frequency analysis, in three problem categories: health, school, and family problems, were used to analyze the characteristics of student suicides. Topic analysis showed that only 30% of the student suicide case reports identified problematic student characteristics related to suicide. Correlations between words showed that words in one problem category were often correlated with words in other problem categories. Frequency word analysis showed that the three problem categories varied across gender and school levels. These results provide interesting insights into the characteristics of suicides among Korean students and important implications for suicide intervention in the education field.