Automatic coherence analysis of Dutch: testing the subjectivity hypothesis on a larger scale

With the increasing availability of large corpora, quantitative corpus analysis is becoming more and more popular as a method for doing linguistic research. This paper uses a new research tool that makes it possible to search syntactically annotated corpora without extensive programming knowledge (CESAR) to study the subjectivity patterns of four Dutch causal connectives. Analyzing a large set of causal relations marked by four of the most frequent Dutch causal connectives (daarom, dus, omdat, and want), the case study aims to corroborate the subjectivity hypothesis established on the basis of smaller scale studies that used manual annotation. The automatic analysis of the subjectivity patterns of Dutch causal connectives illustrates the usability of CESAR in particular and the feasibility of automatic coherence analysis in general. In addition, it generates new insights into the subjectivity patterns of daarom, dus, omdat, and want.