Musing about why spacy vocabulary size improves NER
I was reading the scispacy paper and came across the below performance of the small and large model on a specific biomedical task:
Paper Review: Leveraging unstructured text and knowledge graph embeddings for BioNLP
Today I will be looking at the paper Integrating Graph Contextualized Knowledge into Pre-trained Language Models a biomedical NLP paper about integrating free text and a knowledge graph for entity typing and relation extraction.
More …