Awesome

sist2 NER repository

How to use with sist2

With sist2 v3.0.4+, go to Configuration > Machine learning options > Model repositories and add this line:

https://raw.githubusercontent.com/simon987/sist2-ner-models/main/repo.json

Then reload the page.

docs/example.png

Models

Models with -q8 indicates 8-bit quantization and -q16 16-bit quantization. Smaller models are faster and better suited for low-spec devices. Larger models are more accurate.

1.bert-ner-* General purpose model trained on tner/btc, tner/conll2003 and tner/wnut2017 for 10 epochs.

How to train models:

with Python 3.9

pip install -r requirements.txt
# For GPU support, follow instructions here: https://www.tensorflow.org/install/pip
bash train.sh

Hacking

Feel free to clone and modify the models for your use-case. You can publish the models on GitHub or any static web hosting platform. PRs welcome!