mingjiechen / nlg Goto Github PK

It is important to evaluate using the e2e scripts from here:

https://github.com/tuetschek/e2e-metrics

They use a different implementation of BLEU score.

Could you add a README file to explain how one can run your code?

the code from tf_seq2seq import sequence_loss in CNN.py, but i didnot find tf_seq2seq?

I put the raw_data.pickle file and the unzipped train directory in the same directory as dagger.py, but when I run dagger.py I get the can not load message. Could you check and maybe add the commands?

Could we have the following for each sentence:

• the input MR
• the output of the system
• the gold standard reference
  ?

This would help understand better what the system does, beyond BLEU scores.