Text quality estimation system (BEA12 workshop)
This code implements the method described in:
- Robert Östling and Gintare Grigonyte: Transparent text quality assessment with convolutional neural networks (to be published in the proceedings of The 12th Workshop on Innovative Use of NLP for Building Educational Applications, September 8th, Copenhagen).
The train_rank.py program performs training, use the --help option to see
all possible arguments.
The score_rank_visualize.py script can output color-coded predictions in
HTML or LaTeX.
Due to privacy concerns we are not able to release the full essay dataset used
in our article, but the file essays.scores contains our model's predictions
for the essays. The columns are: predicted score, grade 1 (assigned during
blind re-grading), grade 2 (assigned by student's own teacher), length in
characters of essay.