TY - GEN
T1 - Rule-based system for automatic grammar correction using syntactic n-grams for english Language Learning (L2)
AU - Sidorov, Grigori
AU - Gupta, Anubhav
AU - Tozer, Martin
AU - Catala, Dolors
AU - Catena, Angels
AU - Fuentes, Sandrine
N1 - Publisher Copyright:
© 2013 Association for Computational Linguistics.
PY - 2013
Y1 - 2013
N2 - We describe the system developed for the CoNLL-2013 shared task—automatic English L2 grammar error correction. The system is based on the rule-based approach. It uses very few additional resources: a morphological analyzer and a list of 250 common uncountable nouns, along with the training data provided by the organizers. The system uses the syntactic information available in the training data: this information is represented as syntactic n-grams, i.e. n-grams extracted by following the paths in dependency trees. The system is simple and was developed in a short period of time (1 month). Since it does not employ any additional resources or any sophisticated machine learning methods, it does not achieve high scores (specifically, it has low recall) but could be considered as a baseline system for the task. On the other hand, it shows what can be obtained using a simple rule-based approach and presents a few situations where the rule-based approach can perform better than ML approach.
AB - We describe the system developed for the CoNLL-2013 shared task—automatic English L2 grammar error correction. The system is based on the rule-based approach. It uses very few additional resources: a morphological analyzer and a list of 250 common uncountable nouns, along with the training data provided by the organizers. The system uses the syntactic information available in the training data: this information is represented as syntactic n-grams, i.e. n-grams extracted by following the paths in dependency trees. The system is simple and was developed in a short period of time (1 month). Since it does not employ any additional resources or any sophisticated machine learning methods, it does not achieve high scores (specifically, it has low recall) but could be considered as a baseline system for the task. On the other hand, it shows what can be obtained using a simple rule-based approach and presents a few situations where the rule-based approach can perform better than ML approach.
UR - http://www.scopus.com/inward/record.url?scp=84976450377&partnerID=8YFLogxK
M3 - Contribución a la conferencia
AN - SCOPUS:84976450377
T3 - CoNLL 2013 - 17th Conference on Computational Natural Language Learning, Proceedings of the Shared Task
SP - 96
EP - 101
BT - CoNLL 2013 - 17th Conference on Computational Natural Language Learning, Proceedings of the Shared Task
PB - Association for Computational Linguistics (ACL)
T2 - 17th Conference on Computational Natural Language Learning: Shared Task, CoNLL 2013
Y2 - 8 August 2013 through 9 August 2013
ER -