A Python Implementation of SARI

The original author of SARI.py is Wei Xu
I extend it to a python package. If you use this project in your research, please cite Wei's awesome paper (see below).
The dataset acompanied by the orignal repo was deleted to make the package lightweighted, but you can still find the dataset here https://github.com/cocoxu/simplification

Installation

python setup.py develop

Example Code

from sari.SARI import SARIsent

ssent = "About 95 species are currently accepted ."
csent1 = "About 95 you now get in ."
csent2 = "About 95 species are now agreed ."
csent3 = "About 95 species are currently agreed ."
rsents = ["About 95 species are currently known .", "About 95 species are now accepted .", "95 species are now accepted ."]

print SARIsent(ssent, csent1, rsents)
print SARIsent(ssent, csent2, rsents)
print SARIsent(ssent, csent3, rsents)

Contact: Wei Xu (Ohio State University)

Code, data and trained models from the following paper:

 @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}
 }

Data

./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 8 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

Code

./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.

The Text Simplificaiton System

The text simplification system was implemented into the MT toolkit Joshua Decoder.