/supervised_oie_wrapper

Thin wrapper for the AllenNLP's implementation of supervised open information extraction

Primary LanguagePythonMIT LicenseMIT

Supervised OIE Wrapper

This is thin wrapper over AllenNLP's pretrained Open IE model.

Outputs predictions identical to those in the onlline demo, with batched gpu options.

Install prerequisites

  • AllenNLP
  • docopt
  • tqdm

Use the following to install requirements:

pip install -r requirements.txt

Run on raw sentences

cd src
python run_oie.py --in=path/to/input/file  --batch-size=<batch-size> --out=path/to/output/file [--cuda-device=<cude-device-identifier>]

If --cuda-device is not specified, the model will run on the cpu.

Input format

Raw sentences, each in a new line.

Output Format

Each line pertains to a single OIE extraction:

tokenized sentence <tab> ARG0:.. <tab> V:... <tab> ARG1:...  ...

Example

See example of input and output files in src/example.txt and src/example.oie.