Automatic Coding of Questioning Patterns in Knowledge-building Discourse

Abstract

We propose a novel method for identifying questioning patterns, which are assumed to be one of the essential factors indicating the quality of knowledge-building discourse. The underlying principle of the proposed method is to extract syntactic and sematic information before segmenting the raw data and annotating them according to a multi-layer framework called ACODEA. As a bottom layer of the framework, the “pre-coding” phase makes it possible to translate the raw data into machine-readable and context-independent language, and to make Natural Language Processing tools aware of users’ preferences and underpinning mechanisms of identifying the desired pattern. Explorative but promising evidence is reported toward a more comprehensive perspective by combining qualitative and quantitative methods to analyze the discourse data. Given those findings, we argue in favor of mixed methods of content analysis and they further generated directions for future methodological development and empirical applications.