Closed cwlinghk closed 6 years ago
The spell checker and models are from another repository (https://github.com/nusnlp/smtgec2017)
You have to use the commands in these lines in this script, and also use a dictionary/word list to identify the unknown words: https://github.com/nusnlp/smtgec2017/blob/master/run.sh#L47-L68 if you are to use the spell checker independently.
In the case of using spell checker at the output of the encoder-decoder GEC system (or an SMT-based GEC system), the word list contains all the source words in the training data. Any other word is treated as unknown word.
The paper contains a spell checked result (table 1). May I know if this repository contains the spell checking model? Thanks a lot.