Hello everyone, can you give some tips for my task?
I have context and response where context some dialog and response just response to this dialog so my model need to classify: this response human-generated or machine-generated.
I have tried seq2seq, RNN, and CNN(text classification) but the results were not so good as I wanted.
I know that this question does not relate to the Pytorch, but may be you have some ideas?
RNN works fine for me it gives ~ 80% accuracy on the validation set but there is some way to get higher accuracy(Up to 90%). This data from Hackathon and the size is 7 millions examples