A Free Software
The Automatic Text Response Coder (ReCo) codes (e.g., scores) your text response data; for example:
“The story is about a girl falling into and wandering through a fantasy world.”
Is this a similar response like another one? Is it a correct response? ReCo can answer those questions by, first, categorizing all responses in your data into response types by their semantics, and then, second, by training the model to assign codes to each response type.
ReCo is specialized in short text responses from assessments such as PISA, but can be employed in other contexts as well. The app’s complete version will allow you to construct your own text corpora from Wikipedia, compute tailored semantic spaces, apply automatic spelling correction, adapt the workflow to your needs by switching to R at any time, export the classifier so that it can be used as a service during assessment, choose other test languages, and much, much more.
So Far Just a Prototype
The app’s current version only demonstrates the methodology behind ReCo. While the app’s goal is to enable you to run automatic coding on your own assessment data, with the version you can install currently, you can foremost use the provided demo data set and pre-computed semantic space.
The app’s functionality will be much more comprehensive in the first stable release.
The same is true for the provided demo data. It contains German text data that has been constructed from real empirical data (but can only be considered simulated data itself), which was only possible for text responses with pretty low language diversity. ReCo’s methodology, however, works for much more complex response semantics and a bunch of other languages as well.
If we got you hooked, stay tuned!