CLARIN Café on Computer-assisted Pragmatic Annotation of native and learner Corpora

The corpora DisDir and Ladder consist of elicited speech-acts of cancellation and request in Italian L1 and L2, German, and Colombian Spanish. They have been collected and partially manually annotated for pragmatic research in transcultural pragmatics and second language education. Based on the aforementioned data, the project LadderWeb aims to:

1. Train a machine-learning based software programmed for the automatic annotation of pragmatic categories in requests and cancellation in Italian L1 and L2 and the other aforementioned languages.
2. Annotate part of the corpus with AI support and archive all the elicited data in the ARCHE CLARIN repository.
3. Make the corpus accessible and queryable for learners and practitioners through a web interface.

Unlike previous attempts at pragmatic annotation, the LadderWeb project is based on elicited data that constrain speech acts, control extratextual contexts and allow the detection of implicit information. This makes it possible to circumvent the traditional problems associated with pragmatic annotation and distinguishes the project as a pioneering achievement in the annotation of native and learner corpora.

This presentation will highlight the steps accomplished to collect and (automatically) annotate the corpus as well as develop the code for the automatic annotation. It will also include a demo and best practice recommendations for educators using the web interface.