Fig. 4From: C-Norm: a neural approach to few-shot entity normalizationArchitecture of our Sieve method using our SLFNN and shallow CNN methods. The inputs are the matrix of word embeddings of the tokens of the mentionsBack to article page