Identifying events using computer-assisted text analysis

Abstract

Events as elections, significant changes in laws, but also extreme weather conditions, may affect societal values and, consequently, public opinion. Accordingly, a central assumption for public opinion surveys in that respondents behavior is influenced by significant events. It is therefore necessary to consider the impact of potential events when designing a survey and, whenever possible, to control for these. To support the documentation of such social events, the authors have developed a procedure to identify events using computer-assisted text analysis. Event words are selected and grouped by means of exploratory factor analysis based on a comparison of a large text corpus that forms the reference for a smaller text corpus consisting of media items on significant events. As a result, the factors represent significant events during a specific time period. (author's abstract)