Constructions: a New Unit of Analysis for Corpus-based Discourse Analysis

PACLIC 28 !634 Constructions: a new unit of analysis for corpus-based discourse analysis Samia Touileb Andrew Salway Information Science and Media Studies Uni Research Computing University of Bergen Thormøhlensgt. 55 N-5020 Bergen Norway N-5008 Bergen Norway samia.touileb@gmail.com andrew.salway@uni.no Abstract We propose and assess the novel idea of using automatically induced constructions as a unit of analysis for corpus-based discourse analysis. Automated techniques are needed in order to elucidate important characteristics of corpora for social science research into topics, framing and argument structures. Compared with cur- rent techniques (keywords, n-grams, and collo- cations), constructions capture more linguistic patterning, including some grammatical phe- nomena. Recent advances in natural language processing mean that it is now feasible to auto- matically induce some constructions from large unannotated corpora. In order to assess how well constructions characterise the content of a corpus and how well they elucidate interesting aspects of different discourses, we analysed a corpus of climate change blogs. The utility of constructions for corpus-based discourse analy- sis was compared qualitatively with keywords, n-grams and collocations. We found that the unusually frequent constructions gave interest- ing and different insights into the content of the discourses and enabled better comparison of sub-corpora. 1 Introduction In recent years, with the increasing availability of online text data and computing power, there has been a rapid increase in interest in corpus-based discourse analysis, particularly among social sci- ence researchers. Within social science, discourse analysis is concerned with how societally im- portant issues and opinions are expressed through language, e.g. in news and social media. The scale of the data sets means that automated techniques are essential, at least to give researchers an over- view of the content in a corpus and to elucidate interesting aspects for further investigation. The aim of this paper is to assess the novel idea of using automatically induced constructions for corpus-based discourse analysis. Section 2 pro- vides some background about corpus-based dis- course analysis and discusses some limitations of the automated techniques that are commonly used. It also describes what constructions are and how some constructions can be induced automatically by taking advantage of recent developments in natural language processing. Then in Section 3 we report our investigation into the use of construc- tions for corpus-based discourse analysis. This compared the utility of unusually frequent con- structions with current techniques, based on how they gave insights into the content of a large cor- pus of climate change blogs, and how they eluci- dated interesting phenomena for further investigation. Section 4 summarises our conclu- sions and contributions, and outlines future work. 2 Background In this section we review the use of automated text analysis techniques for corpus-based discourse analysis, and explain why we propose construc- tions as a new unit of analysis (section 2.1). Then we explain how the state-of-the-art in grammar Copyright 2014 by Samia Touileb and Andrew Salway 28th Pacific Asia Conference on Language, Information and Computation pages 634–643 PACLIC 28 !635 induction means that it is now possible to automat- ically …

Content truncated to 500 words