search

freesunshine0316 / neural-graph-to-seq-mp

Code corresponding to our paper "A Graph-to-Sequence Model for AMR-to-Text Generation"
138 stars 26 forks source link

dataset #3

Closed Gitsamshi closed 6 years ago

Gitsamshi commented 6 years ago

I run your model on the corpus of the little prince and the results were quite good. Was little prince corpus already included in the training set? Thank you

freesunshine0316 commented 6 years ago

I don't think so. The only training data is the LDC2015E86 and 2 million raw data sampled from Gigaword, parsed by JAMR.

lurr98 commented 3 years ago

@Gitsamshi do you happen to know the bleu score on the little prince corpus you got? I'm getting a score of 33.25 on the testset but only 13.5 on the little prince corpus, so I'm not sure if something went wrong while preprocessing the corpus