Try transformer based models

[ashridh]

Slightly out of the loop here, but some classifiers we can try out

• Bert
• RoBERTa
• DistilBERT
• GPT

They've shown promising performance on most text based tasks, might be worth looking into them. (it has also become surprisingly convenient to use pre-training with these models nowadays)

[marco-c]

I tried DistilBERT out of curiosity for the components model (using the transformers library) and it looks promising.

[marco-c]

Closing as duplicate of #39.