Stanford NER

A Java-based implementation of Conditional Random Field sequence models for labeling entities like persons, locations, and organizations.

Stanford NER (Named Entity Recognition) provides high-performance tools for identifying rigid designators in text using Linear Chain CRF models. It ships with pre-trained classifiers for English (3-class, 4-class, and 7-class models), Chinese, and German, while allowing users to train custom models on their own labeled data. Developed by the Stanford Natural Language Processing Group, this software integrates seamlessly into the broader Stanford CoreNLP pipeline and remains a foundational standard for extracting structured data from unstructured strings.

https://nlp.stanford.edu/software/CRF-NER.shtml