MITIE

Why we included this project

MITIE offers a way to pull named entities and binary relations out of raw text without leaning on a hosted API. The core is a C++ library built on dlib that combines distributional word embeddings and structural support vector machines for extraction. Pre-trained models ship for English, Spanish, and German, trained on resources like CoNLL 2003, ACE, Wikipedia, and Freebase, so you can get going without training your own. Bindings for Python, R, Java, C, and MATLAB make integration straightforward, and community bindings extend that to OCaml, .NET, PHP, and Ruby. If you need to label your own data, the example programs show how to train custom extractors and relation detectors. MITIE is licensed for commercial use and has been around a while, so expect to build it yourself and manage model downloads rather than get frequent updates.