ScispaCy: Fast and Robust Models for Biomedical Natural Language Processing

About

Despite recent advances in natural language processing, many statistical models for processing text perform extremely poorly under domain shift. Processing biomedical and clinical text is a critically important application area of natural language processing, for which there are few robust, practical, publicly available models. This paper describes scispaCy, a new tool for practical biomedical/scientific text processing, which heavily leverages the spaCy library. We detail the performance of two packages of models released in scispaCy and demonstrate their robustness on several tasks and datasets. Models and code are available at https://allenai.github.io/scispacy/

Mark Neumann, Daniel King, Iz Beltagy, Waleed Ammar · 2019

Related benchmarks

Task	Dataset	Result
Entity Linking	MM-ST21PV English	Recall@153.8	13
Entity Linking	QUAERO-EMEA French	Recall@10.371	13
Entity Linking	SympTEMIST Spanish	Recall@19.8	13
Entity Linking	DisTEMIST Spanish	Recall@121.1	13
Entity Linking	MedProcNER Spanish	Recall@110.3	13
Entity Linking	MM-ST21PV english (test)	Recall@153.8	11
Entity Linking	QUAERO-MEDLINE french (test)	Recall@140.5	11
Entity Linking	QUAERO-EMEA french (test)	Recall@137.1	11
Entity Linking	SPACCC spanish (test)	Recall@113.2	11
Named Entity Recognition	NBCI-Disease preprocessed (test)	Micro F1 (Excl. O)81.65	4