Fig. 3From: BioByGANS: biomedical named entity recognition by fusing contextual and syntactic features through graph attention network in node classification frameworkAn example of using the one-hot strategy to encode syntactic features. In this example, dyskinesia, which is a named entity, is a “PROPN” in the sentence, and has 2 dependencies with other words. Moreover, the tokenizer of BioBERT cuts dyskinesia into 5 word-pieces, which inherit the syntactic features of the original word in the sentence, and are linked by “next” to indicate the order of word-piecesBack to article page