georgebrock / ner-tools

Tools for training Stanford NLP's NER models

Stanford Named Entity Recognition

Tools for training a Stanford NER model, based on the example in the Stanford NER FAQ.

Setup

0. Install Java.
1. Download Stanford NER from https://nlp.stanford.edu/software/CRF-NER.shtml#Download
2. Extract it to a directory named stanford-ner in this directory.

Usage

1. Provide two sets of examples in train.txt and test.txt with one example on each line.
2. Run make sources to produce train.tsv and test.tsv.
3. Edit train.tsv and test.tsv, replacing O with entity names where appropriate. This may take some time.
4. Run make model to produce the model, in ner-model.ser.gz.
5. Run make check to test the model against the examples from test.txt.