From: A semi-supervised short text sentiment classification method based on improved Bert model from unlabelled data
Corpus
Instance
Average word
Total word
Partition
Amazon Reviews training
58,000
26.187
1,518,875
Training (58,000)
Amazon Reviews verification
7250
24.823
179,965
Verification (7250)
Amazon Reviews testing
25.156
182,383
Testing (7250)