Closed FatemehMashhadi closed 2 years ago
Hi @FatemehMashhadi, we've added the code for reference only.
I've added the main script here: https://github.com/martiansideofthemoon/style-transfer-paraphrase/blob/master/datasets/parse_paranmt_postprocess.py The command to run it is on top of the script (don't use default values).
This code operates on cached data, which was generated using the get_kendall_tau() and f1_score functions here: https://github.com/martiansideofthemoon/style-transfer-paraphrase/blob/master/datasets/preprocess_utils.py. You will need to write a script to run it (it was sort of slow so I had parallelized it on the cluster). The scripts for paraphrase similarity can be found here: https://github.com/martiansideofthemoon/style-transfer-paraphrase/tree/master/style_paraphrase/evaluation#similarity
If you need the cached data as well, please drop me an email at kalpesh@cs.umass.edu
Is there any pipeline code to filter the PARANMT-50M corpus?