Poor Bleu Score

Hi there,

I have run the OpenNMT-py model using my own dataset (constructed similar to that of the provided one) but I am getting very poor Bleu scores. At first, it looks like a case of overfitting for which I have provided different dropout probabilities. But final results are still not very good. May I request you to suggest me what should I do? I have seen that the equations are not getting properly predicted (either full or partial equations are getting incorrectly predicted).

Thanks in advance!