Contact: Wei Xu (Ohio State University)
Code, data and trained models from the following papers:
@article{Xu-EtAl:2016:TACL,
author = {Wei Xu and Courtney Napoles and Ellie Pavlick and Quanze Chen and Chris Callison-Burch},
title = {Optimizing Statistical Machine Translation for Text Simplification},
journal = {Transactions of the Association for Computational Linguistics},
volume = {4},
year = {2016},
url = {https://cocoxu.github.io/publications/tacl2016-smt-simplification.pdf},
pages = {401--415}
}
and
@article{Xu-EtAl:2015:TACL,
author = {Wei Xu and Chris Callison-Burch and Courtney Napoles},
title = {Problems in Current Text Simplification Research: New Data Can Help},
journal = {Transactions of the Association for Computational Linguistics},
volume = {3},
year = {2015},
url = {http://www.cis.upenn.edu/~ccb/publications/publications/new-data-for-text-simplification.pdf},
pages = {283--297}
}
./tacl2016-smt-simplification.pdf the paper
./data/turkcorpus/ tuning and test data
*.norm tokenized sentences from English Wikipedia
*.simp tokenized, corresponding sentences from Simple English Wikipedia
*.turk.0~7 8 reference simplifications by different Amazon Mechanical Turkers
./data/systemoutputs/ 4 different system outputs compared in the paper
./data/ppdb/ppdb-1.0-xl-all-simp.gz (a 3.8G file) paraphrase rules (PPDB 1.0) with added simplification-specific features
./data/ppdb/ppdb-1.0-xxxl-lexical-self-simp.gz (a 27M file) self-paraphrase lexical rules that map words to themselves, and help to copy input words into outputs
./SARI.py a stand-alone Python implementation of the SARI metric for text simplification evaluation
There is also a Java implementation of SARI that is integrated as part of the Joshua's codebase.
./HIT_MTurk_crowdsourcing/ HTML interfaces designed for human evaluation of simplification systems, as well as parallel corpus collection (originally used on Amazon Mechnical Turk HITs)
The text simplification system was implemented into the MT toolkit Joshua Decoder.
./ppdb-simplification-release-joshua5.0.zip (a 281M file) The experiments in our TACL 2016 paper used the Joshua 5.0. Example scripts for training the simplification are under the directory ./bin/. The joshua_TACL2016.config is also provided -- that is corresponding to the best system in our paper. You may find the Joshua pipeline tutorial useful. Note that STAR is corpus-level version of SARI, SARI is sentence-level; the current STAR.java used the hardcoded the number of reference sentences to 8, and used F_score of the deletion rather than only the precision (you may want to change before using it).
./scripts_preprocessing/ The tokenizer and sentence spliter used for preprocessing.