"Automatic" Analysis
Several quantitative studies have attempted to extract frames from textual data through clustering methods, such as hierarchical cluster analysis or factor analysis.
The best known algorithm to detect frames quasi automatically via the keyword approach is implemented in Miller's (1997) VBPro program. VBPro was specifically developed for frame analyses and detects frames via hierarchical cluster analysis using data in plain text format. The program requires roughly equally big chunks of data. The program has been successfully tested in a number of empirical studies.
However, other word clustering techniques might be suited better for frame analyses, even though they were not specifically designed for it. For instance, the Leximancer algorithm contain several advantages in addition to its more intuitive interface. Unlike VBPro, it does not require the use of stop words, which indeed can be of utmost importance for the interpretation of results. Our own tests confirm research that has shown that removal of homonyms and deictics in unsupervised word clustering might decrease, rather than increase the validity of results (Burstein et al. : 5; Smith & Humphreys, 2006). Leximancer also automatically creates disambiguation rules and synonym lists, which further the validity of results (Smith, 2000: 2).
A number of other clustering and text mapping algorithms, among them Alceste, AutoMap, Hamlet, Tropes, T-LAB, and Sphinx Survey Lexica, exists, but unfortunately to date, there has been no comparative evaluation, as to which of these programs yields the most useful results. In fact, most programs have not even been examined on the basis of traditional validity criteria for content analysis (Smith & Humphreys, 2006).
Qualitative Programs
Besides these automatic frame extraction programs, CAQDAS can aide more qualitative oriented framing studies in the coding of data. Only a handful of framing studies to date have used these one of programs, namely ATLAS.ti (Trenz, 2004), MAXqda/winMAX (Downey & Koenig, 2006) ; van de Steeg et al., 2003), and NUD*IST (Reese et al.)Of the major CAQDAS, MAXqda and QDA Miner have been shown to be most suitable for the use in mixed-methods approaches to framing (Koenig, 2005).
References
- Burstein, Jill, Daniel Marcu, Slava Andreyev, and Chodorov, Martin. Towards Automatic Classification of Discourse Elements in Essays. ETS. http://www.ets.org/research/dload/burstein.pdf.
- Downey, John W. and Thomas Koenig. 2006. Nationalization vs. Europeanization vs. Globalization of Issues that Should Belong to the European Public Sphere. European Societies 6 (3).
- Koenig, Thomas 2005. "Routinizing Frame Analysis", Proceedings of the ISA RC-33 Methodology Conference, Leverkusen: Leske & Budrich.
- Reese, Stephen D., Oscar H. Gandy, and August E. Grant. 2001. Framing Public Life: Perspectives on Media and Our Understanding of the Social World. Mahwah, NJ: Lawrence Erlbaum Associates.
- Smith, Andrew E. 2000. Machine Mapping of Document Collections: The Leximancer System. Proceedings of the Fifth Australasian Document Computing Symposium. Melbourne, Australia.
- Smith Andrew E. and Michael S. Humphreys. 2006. Evaluation of unsupervised semantic mapping of natural language with leximancer concept mapping. Behavior Research Methods 38.
- Trenz, Hans-Jörg. 2004. Media Coverage on European Governance: Exploring the European Public Sphere in National Quality Newspapers. European Journal of Communication 19 (3) 291-319.
- van de Steeg, Marianne, Valentin Rauer, Sylvain Rivet, and Thomas Risse. 2003. "The EU As a Political Community: A Media Analysis of the 'Haider Debate' in the European Union." Paper Presented at the Annual Meeting of the European Union Studies Association (EUSA). Nashville, TN, March 27-30, 2003. <http://www.fu-berlin.de/atasp/texte/030322_eu_pol.pdf>