1- Carry out manual error analysis to try to understand the difference between training and dev/test sets.

2- Make training data more similar or collect more data similar to dev/test sets.

3- Merge a subset of the dev/test set into the training set and use smaller dev/test sets. It's not a good idea to merge a subset of the training set into dev/test sets.

4- Training-dev set: a validation set with the same distribution as training set, but not used for training.