Discourse Network Analyzer (DNA)
Overview
DNA is a Java-based software for qualitative category-based content analysis. It serves two purposes: coding statements of actors into categories, and converting these structured data into networks that are readable by UCINET, visone and other network-analytic software packages. The software can extract bipartite (affiliation) networks, co-occurrence networks of actors or concepts, and dynamic animations of these networks.
Full control due to manual coding
The Discourse Network Analyzer follows a very simple and intuitive logic: you insert text documents from any source (via copy & paste), then you go through the text and highlight statements of actors with your mouse.
You encode each statement with four variables:
- either the person or the organization (or both) who is making the statement,
- the category the actor is referring to (this can be a solution concept, a causal perception or whatever your actors talk about),
- and a dummy (yes/no) variable indicating the actor's agreement or disagreement with the aforementioned category.
There are many convenience functions to assist you in navigating through your actors' statements, filtering them, recoding them and querying them for actors' self-contradictions. All previous actors and categories are saved in a drop-down box in order to avoid typos.
Automatic coding is currently not supported.
Network export functions
The main goal of the Discourse Network Analyzer is to visualize a (political, cultural or whatever) discourse as a network. That means you can export your structured data to a matrix or edge list. The .dl, .graphml and .csv files can be read by most network-analytic software packages, including Ucinet, visone, R and spreadsheet software like MS Excel or OpenOffice. DNA is the bridge between content analysis (which is done in DNA) and the visualization of the resulting networks (which is done in network analysis software like Netdraw or visone).
DNA offers a variety of options for exporting your data as networks (see the screenshots for details).
What kinds of networks can DNA export?
The most simple kind of network is an affiliation graph, where actors are connected to the concepts they advocate or reject (this can be visualized by different edge colors, for instance). However, this kind of network diagram is fairly complex and hard to grasp. For this reason, other kinds of networks have been developed:
Actor congruence networks show only persons or only organizations. Actors are connected by a tie if they have at least one concept in common (both of them agreeing to this concept or both of them rejecting the concept). The edge weight, which can be visualized by the line width, is proportional to the number of the concepts they have in common. Clusters in this network correspond to discourse coalitions, belief coalitions, advocacy coalitions, epistemic communities or other constructs (depending on the type of statement encoded).
The same kind of co-occurrence network can be exported for concepts. The concept congruence network can be used as a map of solution concepts or causal beliefs. An actor conflict network can be created where two actors are connected if one of them likes a concept and the other actor dislikes the concept or vice-versa.
Finally, several kinds of dynamic networks can be exported, where the development of the discourse over time can be observed. For example, it is possible to watch an animation of how a discourse evolves over time. How and when do clusters or coalitions in the discourse emerge, and to what extent to actors leave their discourse coalitions and join other coalitions over time?
Controlling DNA from R
The Discourse Network Analyzer can be used within the statistical programming environment R. It is fairly easy to import networks directly from DNA into R as matrices. If you change some of your codes in your .dna file later on, you can just re-run your R script and don't have to work through all the network export options again. To use DNA with R, you can use the R package rDNA.
Other features
DNA is also able to export time series statistics of actors' statement frequencies per month or per year (see screenshot 3), and it has advanced search functions and keyword highlighting functions to facilitate the coding of statements in the text.
Moreover, DNA is based on UTF-8 Unicode character encoding and can thus be employed for text sources in any non-Western language.
DNA is currently based on an XML file format, which means that the .dna data files can be opened in any text editor (in case you would like to re-use your data in another program later on).
Is DNA free to use? How can I get started?
DNA can be downloaded and used for free. The developer does not assume any liability though. DNA is not open-source, but it can be used free of charge.
The application was developed, tested and heavily used in a dissertation research project on German pension politics. A copy of the dissertation can be requested by e-mail. The results are currently being published in academic journals mainly related to political science (see the publications page for details). To learn more about DNA, you may want to read the manual or my recent EJPR article.
Help or feedback is provided on the DNA-help mailing list. Some DNA users and research projects based on DNA are listed on the DNA users page. You can obtain a free copy of DNA from the download page.