Data and source code accompanying the paper "Split and Rephrase: Better Evaluation and a Stronger Baseline".

https://arxiv.org/abs/1805.01035

Roee Aharoni and Yoav Goldberg, ACL 2018

The data and some of the scripts are based on the repository by Narayan et al.: https://github.com/shashiongithub/Split-and-Rephrase

This repository includes:

• The proposed data split, under data/baseline-seq2seq-split-RDFs-relations.zip.

• The original data split, under data/baseline-seq2seq.zip.

• The raw sentences and .id files required in the test.py and validate.py scripts, under data/complex-sents.zip.

  • Additional files required by the test.py and validate.py scripts of our models (i.e. under the benchmark and evaluation-directories... folders are available here.

• Scripts for:

  • Training our proposed models using openNMT-py (under src/training_scripts)

  • Evaluating the models as proposed by Narayan et al., 2017 (under src/evaluate.py)

  • Creating the RDF-based data split to reduce overlap between the development and test set found in the original split (under src/data/create_new_split.py)

Prerequisits:

• PyTorch

• openNMT-py (commit d4ab35a)

Feel free to reach out in roee.aharoni@gmail.com if you have any further questions!