First of all, thanks for your nice work and code.
I have several questions on the paper and code.
Can you elaborate more on measuring with BLEU-4? BLEU has many smoothing methods and measuring ways. Also many packages and codes. But I don't see any BLEU on this repo nor Graph2Seq repo. I want to try it myself so I'm asking this question.
In the paper, Seq2Seq baseline has a template. How exactly and is dealt with? Take pretrained word embedding of word 'max' for MAX?
I think there are a lot of OOV words in WikiSQL dataset, expecially in condition values. If you use pretrained GloVe, then how did you handle that?
First of all, thanks for your nice work and code. I have several questions on the paper and code.
Can you elaborate more on measuring with BLEU-4? BLEU has many smoothing methods and measuring ways. Also many packages and codes. But I don't see any BLEU on this repo nor Graph2Seq repo. I want to try it myself so I'm asking this question.
In the paper, Seq2Seq baseline has a template. How exactly and is dealt with? Take pretrained word embedding of word 'max' for MAX?
I think there are a lot of OOV words in WikiSQL dataset, expecially in condition values. If you use pretrained GloVe, then how did you handle that?
Thank you.