the experiments setting for 20news
Opened this issue · 1 comments
xiaoheizai commented
Dear leishu:
When I reproduce the code using pytorch, I find that the result of 20news is lower under the 25% setting. I guess it is caused by inconsistent data processing，so I hope to know the details of the data processing of 20news.
Thank you !
leishu02 commented
（1）use every word inside the document.
（2）build vocabulary (pre-trained embedding) based on both training and testing data.
Please refer to the code regarding other settings.
Dear leishu:
When I reproduce the code using pytorch, I find that the result of 20news is lower under the 25% setting. I guess it is caused by inconsistent data processing，so I hope to know the details of the data processing of 20news.
Thank you !