Developing Methodology for Korean Particle Error Detection
Markus Dickinson, Ross Israel, Sun-Hee Lee · 2011
We further work on detecting errors in postpositional particle usage by learners of Korean by improving the training data and developing a complete pipeline of particle selection. We improve the data by filtering non-Korean data and sampling instances to better match the particle distribution. Our evaluation shows that, while the data selection is effective, there is much work to be done with preprocessing and system optimization. 1