Emoji Prediction Beyond the Text via BiLSTM with Label-wise Attention and BERT
Data Size : 402k English tweets for training, 50k for validation, and
50k for testing.
Topic : Natural Language Processing
Extract the users’ emotions from
their tweets based on the tweet’s features
such as the text, hashtag, and source to
predict the relevant emoji instead of using
rule-based NLP techniques as most devices
and applications are using these days.