Colaboratory notebooks

• FSPD_transformer.ipynb: using a custom transformer architecture to classify food policies by their description (see paper [linktbd]) Own tokenization and embedding layers within the model. 2-headed attention. 75% accuracy on test set.

• FSPD_MLP.ipynb : using a MLP to classify food policies by their description.(see paper [linktbd]) Tokenizer & embeddings: distilbert-base-uncased. ~90% accuracy on test set.

• FSPD_lgb.ipynb : using a lightgbm to classify food policies by their description.(see paper [linktbd]) Tokenizer & embeddings: distilbert-base-uncased. ~95% accuracy on test set.

• Gensim_Lovecraft.ipynb: playing around with Lovecraft's corpus using Gensim and Word2Vec.