| Resumen |
In this paper, we present a system for automatic English (L2)
grammatical error correction. It participated in ConLL 2013
shared tasks. The system applies a set of simple rules for
correction of grammatical errors. In some cases, it uses
syntactic n-grams, i.e., n-grams that are constructed in a
syntactic metric: namely, by following paths in dependency
trees, i.e., there is special procedure that allows obtaining
syntactic n-grams. Note that in general case syntactic n-grams
permit introducing syntactic information into machine learning
methods, because syntactic n-grams have all properties of
traditional n-grams. The system is simple, practically does not
use additional linguistic resources and was constructed in two
months. Due to its simplicity it does not obtain better scores as
compared to more sophisticated systems that use many
resources, the Internet and machine learning methods, but it
can be positioned as a baseline system for the task. |
