The results are not consistent with the paper. Could you provide the full datasets and parameters.

wangliangguo / BiLSTM_ILP

4 stars 1 forks source link

The results are not consistent with the paper. Could you provide the full datasets and parameters. #1

Open donghaoye opened 6 years ago

ShirleyHan6 commented 5 years ago

Hi, I am also trying to use this model.

Have you been able to solve this problem, or have you been able to get any other sentence compression model to work?

Thanks. Shirley

wangliangguo commented 5 years ago

Hi, I am sorry that I haven't logged in this account for a long time. For the full datasets, they could be downloaded from google news and NBC News.

The preprocess is to compare the source sentence and compressed sentence, if the word appear in the compressed sentence, the label for the source word is 1 else 0.

This work was done 2 years ago when I was a visiting student in Singapore manage university, so I am sorry that I could not find the pretrained model and whole processed datasets.