This is text error detection code using BiLSTM.
The shape of input data is as below
0\t0\t0\t1\t0\t'Thanks for your attentions .'
The left side (index 0 ~ -2) is labels which mean correct or incorrect.
The right side (index -1) is the sentence splited by spaces.
In this repository, I used FCE Dataset for error detection
By parsing this dataset through parse.py
, you can get the corpus for train/dev/test.