Analysing Journalistic Discourse And Finding Opinions Semi-Automatically?: A Case Study on the 2007 and 2012 Presidential French Campaigns

Abstract

The goal of this paper is to test three different NLP technologies to analyze French journalistic discourse about candidates during French presidential campaigns in order to evaluate discourses differences depending on the candidate gender and/ or the candidate political party. Indeed it is suggested that working on a journalistic corpus with specific software can help studying linguistic patterns and choices which are made on the basis of political affiliation or gender stereotypes. These conclusions are drawn from quantitative and qualitative analysis carried out with 1. the software SEMY which gives semantic profiles semi-automatically

2. the software ANTCONC which provides useful Keyword in Context (KWIC) or abstracts of the text in which is used the studied item, as well as collocations

3. the software TERMOSTAT which works on discourse specificities, frequencies and most used morpho-syntactic patterns. Convergent asymmetries between female and male candidates in journalistic discourse (however conditionally) were found as far as our data dedicated to the 2007 and the 2012 presidential campaigns are concerned. We conclude that social gender (i.e. stereotypical expectations about who will be a typical member of a given category) and / or political favoritism may affect the representation of leadership in discourse and may affect in turn the readership, hence the electorate. Thus the paper recommends the use of corpus linguistic tools to support semi-automatic investigation of characteristics of political texts.