| # Span Finder (v0.0.2) | |
| ## Installation | |
| Environment: | |
| - python >= 3.7 | |
| - pip | |
| To install the dependencies, execute | |
| ``` shell script | |
| pip install -r requirements.txt | |
| pip uninstall -y dataclasses | |
| ``` | |
| Then install SFTP (Span Finding - Transductive Parsing) package: | |
| ``` shell script | |
| python setup.py install | |
| ``` | |
| ## Prediction | |
| If you use SpanFinder only for inference, please read [this example](scripts/predict_span.py). | |
| ## Demo | |
| A demo (combined with Patrick's coref model) is [here](https://nlp.jhu.edu/demos/lome). | |
| ## Pre-Trained Models | |
| Some parameters trained on FrameNet can be found at CLSP grid: `/home/gqin2/public/release/sftp/0.0.2`. | |
| The file `model.tar.gz` is pointed to the best model so for, and file named after dates will not be updated anymore. | |