A live instance of the app is accessible at https://jupyter.korpus.cz/shiny/lukes/mda/.
MDA is a data-driven approach to analyzing register variation in language. A variety of linguistic features are identified in a corpus of diversified texts and their patterns of co-occurrence are then summarized using factor analysis (or some other statistical procedure aimed at reduction of dimensionality). Dimensions of variation in the corpus of texts are thus identified based on text-internal criteria and subsequently interpreted w.r.t. which linguistic features they correlate with.
This Shiny app facilitates visual exploration of MDA results, which is intended to help with the interpretation task.
For a thorough introduction to MDA, see Douglas Biber (1988), Variation across speech and writing (Cambridge: Cambridge University Press).
This is a fairly standard Shiny app, you just need to give it the factor
analysis results as *.RData files under the ./results directory. These
files are expected to contain a factors variable (with the scores of the
individual texts in the corpus w.r.t. the different factors), and a load
variable, which is a matrix of the loadings of the individual linguistic
features on the factors (as retrieved from the factor analysis).
I don't suspect this app will be useful outside our project, but don't hesitate to let me know if you feel otherwise and I'll fill in the details :)
Copyright © 2017--present ÚČNK/David Lukeš, Václav Cvrček
Distributed under the GNU General Public License v3