Make rnn (or other) model to beat naïve bayes baseline

[marksverdhei]

Turned out that the naive bayes model is still better than the rnn. This is unacceptable, and therefore we must tune hyperparameters and experiment so that it can outperform the baseline

[marksverdhei]

Some approaches to upgrade or replace our current neural model:

• Use pre-trained static embeddings
  • Word2vec
  • FastText
  • Glove
• Use contextual embeddings
  • NorElmo
  • NorBert

Or simple replace the rnn with a classification layer on the contextual models. Apart from that, tune hyperparams using the new conf setup