Home

Awesome

AGRR-2019 Gapping Resolver

This is a system submitted to Dialog Evaluation 2019 gapping resolution track.

Description

This system uses an AWD-LSTM encoder to create sentence and token representations and runs an MLP classifier and a linear decoder to find sentences with gapping and attempt to resolve it, respectively.

Metrics

CorpusBinaryResolution
TrainTBDTBD
DevTBDTBD
TestTBDTBD

Usage

To run the experiment:

  1. Install the requirements from requirements.txt.
  2. Fetch folders artifacts and data from Google Drive and put it in the root folder of the repostiroty
  3. Run python resolver.py input_file.csv output_file.csv. The input file should follow the AGRR format.