UD as an annotation standard for learner language: from guidelines to applications

Summary

The project, with the preliminary title UD as an annotation standard for learner language: from guidelines to applications revolves around leveraging the Universal Dependencies (UD) framework to develop cross-lingually applicable approaches to Learner Corpus Research (LCR) studies and Computer-Assisted Language Learning (CALL). Its main contributions are:

• a review of annotation practices across learner treebank (see paper 1 below)
• a proposal for shared annotation guidelines for learner language (also discussed in paper 1)
• a gold-standard treebank of L2 Swedish (paper 2)
• a series of experiments on the automatic parsing of learner language (papers 2, 3 and 4)
• a search tool for parallel (learner) treebanks (paper 5)
• a prototype automatic error analysis tool (paper 6).

Publications

The thesis is planned to include the following publications:

1. Arianna Masciolini, Aleksandrs Berdicevskis, Maria Irena Szawerna, and Elena Volodina. Annotating second language in Universal Dependencies: a review of current practices and directions for harmonized guidelines. In Proceedings of the Eighth Workshop on Universal Dependencies @ SyntaxFest, 2025
2. [preprint] Arianna Masciolini, Maria Irena Szawerna, Aleksandrs Berdicevskis, Caroline Grand-Clement and Elena Volodina. A parallel treebank of learner Swedish
3. Arianna Masciolini, Emilie Francis and Maria Irena Szawerna. Synthetic Error-Augmented Parsing of Swedish as a Second Language: Experiments with Word Order. In Proceedings of the Joint Workshop on Multiword Expressions and Universal Dependencies (MWE-UD) @ LREC-COLING, 2024
4. [in-progress] Arianna Masciolini and Aleksandrs Berdicevskis. Effects of learner data on parser training (working title)
5. Arianna Masciolini, Herbert Lange and Márton András Tóth. Exploring parallel corpora with STUnD: A Search Tool for Universal Dependencies. In Huminfra handbook: Empowering digital and experimental humanities, 2025
6. Arianna Masciolini, Elena Volodina, and Dana Dannélls. Towards automatically extracting morphosyntactical error patterns from L1-L2 parallel dependency treebanks. In Proceedings of the 18th Workshop on Innovative Use of NLP for Building Educational Applications (BEA), 2023

Seminars

1. ideas seminar: UD-based analysis of grammatical errors in L2 texts (April 3, 2023)
2. halfway seminar: Cross-lingual approaches to computational SLA: the potential of Universal Dependencies (October 21, 2024)
3. final seminar: planned for September 14, 2026 (opponent: Sara Stymne)