id	author	title	date	pages	extension	mime	words	sentences	flesch	summary	cache	txt
work_pdhkjvfkbnb4ngi7bpbilobcq4	Alla Rozovskaya	Building a State-of-the-Art Grammatical Error Correction System	2014	16	.pdf	application/pdf	9663	962	68	participated in the task developed a wide array of approaches that include discriminative classifiers, language models, statistical machine-translation systems, and rule-based modules. linguistic knowledge when developing error correction modules, e.g., to identify which type of verb NB model trained on Web1T and adapted to learner errors There are different ways to adapt a model that depend on the type of training data (learner or native) With adaptation, models trained on native data can use the author's word (the source word) as a feature and thus Adaptation always helps on the CoNLL training data and the FCE data (except noun errors), but trained on learner data with word n-gram features and the source better approach is to train on learner data for agreement mistakes and on native data for form errors. the NTHU system also corrects all verb errors using a model trained on Web1T but handles all these	./cache/work_pdhkjvfkbnb4ngi7bpbilobcq4.pdf	./txt/work_pdhkjvfkbnb4ngi7bpbilobcq4.txt