Stanza

Stanza is the Stanford NLP Group's official Python library for multilingual natural language processing. It provides pretrained neural pipelines for tokenization, multi-word token expansion, part of speech tagging, lemmatization, dependency parsing, named entity recognition, and sentiment analysis, with models trained on Universal Dependencies treebanks covering more than 60 human languages. Built on PyTorch, it supports GPU acceleration and batch processing, ships specialized biomedical and clinical English models for scientific literature and clinical notes, and can train custom models from CoNLL-U or BIOES formatted data. The library also acts as a Python client for the Java Stanford CoreNLP server, giving access to that toolkit's additional functionality. As the successor to CoreNLP for Python users, it is a common choice for researchers and practitioners doing syntactic analysis or entity extraction across many languages. Released under the Apache 2.0 license.