Does a modal verb express possibility, permission or capability?
Ana Marasovic and Anette Frank. (2016): Multilingual Modal Sense Classification using a Convolutional Neural Network. Proceedings of the 1st Workshop on Representation Learning for NLP, Berlin, Germany, pp. 111--120, Association for Computational Linguistics.
Internally we used following abbreviations,
- balance, unbalance - dataset was balanced or not
- classifier1, classifier2, classifier3 - MPQA, EPOS+MPQA, EPOS Please refer to the paper what MPQA, EPOS+MPQA, EPOS are.
Not everything is provided with the repo. More precisely,
- there is no code for loading dependency-based embeddings (trivial)
- words which don't have a pre-trained embedding are not randomly initializing with a random vector sampled from uniform distribution which has the same variance as the pre-trained embeddings (trivial)
- there is no code for micro-average calculation (trivial)
- no code for baselines
- no code for significance testing (trivial)
- no code for semantic feature detectors
- no code for WSD
The code is re-factorized after the submission and results could differ from the reported results. The script how to get individual results and macro-average can be found in bash_scripts/epos_eng.sh. If you get results that differ in large margins, please contact me.
The code is hard-coded with paths to the English datasets, change paths for experiments on German data.
If you make use of the contents of this repository, please cite the following paper:
@inproceedings{marasovic2016msc,
title={{Multilingual Modal Sense Classification using a Convolutional Neural Network}},
author={Marasovi\'{c}, Ana and Frank, Anette},
booktitle={Proceedings of the 1st Workshop on Representation Learning for NLP},
year={2016}
}