L2 Finnish model is a classification model trained (finetuned) with data (ICLFI, LAS2, CEFLING, and TOPLING) containing fictional and non-fictional texts written by Finnish as a second language (L2) speakers. With the model you can classify texts into the following CEFR classes: A1, A2, B1, B2, and C.
See the jupyter notebook file (.ipynb) in the Files and versions page for a tutorial.
The preprocessing scripts used for ICLFI, LAS2, CEFLING, and TOPLING have been released in Github:
https://github.com/idatoivanen/finnish_cefr_preprocessingMore information about training data and model training can be found in the paper referenced below (or
here).
1@article{tarvainen2025automatic,
2year = {2025},
3author = {Tarvainen, Jenny and Toivanen, Ida and Huhta, Ari},
4title = {Automatic language proficiency assessment of written texts: Training a CEFR classifier in L2 Finnish},
5journal = {Studies in Language Assessment},
6volume = {14},
7issue = {2},
8pages = {58-90},
9doi = {10.58379/YWAV5140},
10url={https://doi.org/10.58379/YWAV5140}
11}
This repository has been produced as part of the
FIN-CLARIAH infrastructure project.
