Читать книгу An Introduction to Text Mining - Gabe Ignatow - Страница 17

CDA involves seeking the presence of features from other discourses in the text or discourse to be analyzed. CDA is based on Fairclough’s (1995) concept of “intertextuality,” which is the idea that people appropriate from discourses circulating in their social space whenever they speak or write. In CDA, ordinary everyday speaking and writing are understood to involve selecting and combining elements from dominant discourses.

While the term discourse generally refers to all practices of writing and talking, in CDA discourses are understood as ways of writing and talking that “rule out” and “rule in” ways of constructing knowledge about topics. In other words, discourses “do not just describe things; they do things” (Potter & Wetherell, 1987, p. 6) through the way they make sense of the world for its inhabitants (Fairclough, 1992; van Dijk, 1993).

Discourses cannot be studied directly but can be explored by examining the texts that constitute them (Fairclough, 1992; Parker, 1992). In this way, texts can be analyzed as fragments of discourses that reflect and project ideological domination by powerful groups in society. But texts can also be considered a potential mechanism of liberation when they are produced by the critical analyst who reveals mechanisms of ideological domination in them in an attempt to overcome or eliminate them.

Although CDA has generally employed strictly interpretive methods, use of quantitative and statistical techniques is not a novel practice (Krishnamurthy, 1996; Stubbs, 1994), and the use of software to create, manage, and analyze large collections of texts appears to be increasingly popular (Baker et al., 2008; Koller & Mautner, 2004; O’Halloran & Coffin, 2004).

A 2014 study by Bednarek and Caple exemplifies the use of statistical techniques in CDA. Bednarek and Caple introduced the concept of “news values” to CDA of news media and illustrated their approach with two case studies using the same collection of British news discourse. Their texts included 100 news stories (about 70,000 words total) from 2003 covering 10 topics from 10 different national newspapers, including five quality papers and five tabloids. The analysis proceeded through analysis of word frequency of the top 100 most frequently used words and two-word clusters (bigrams), focusing on words that represent news values such as eliteness, superlativeness, proximity, negativity, timeliness, personalization, and novelty. The authors concluded that their case studies demonstrated that corpus linguistic techniques (see Appendix F) can identify discursive devices that are repeatedly used in news discourse to construct and perpetuate an ideology of newsworthiness.

In another CDA study, Baker and his colleagues (2008) analyzed a 140-million-word corpus of British news articles about refugees, asylum seekers, immigrants, and migrants. They used collocation and concordance analysis (see Appendix F) to identify common categories of representation of refugees, asylum seekers, immigrants, and migrants. They also discussed how collocation and concordance analysis can be used to direct researchers to representative texts in order to carry out qualitative analysis.

Research in the Spotlight