I am going to build a dataset for a Sentiment Analysis job. The Positive and Negative statements(tweets) about the specific topic (which is Vaccination) can almost be specified by both human and model. However, there are Neutral statements they are difficult to be labeled. The Neutral ones can be so vast and there is no any labeled-dataset in this ground.
How can I prepare Neutral records for my dataset to be able to use it next in BERT model?
1 Like
up discussion
1 Like