Datasets:

Modalities:
Text
ArXiv:
Tags:

How to obtain duplicates from minhash?

#8
by cq - opened

Hi, thanks for your great job.
I have tried to use the duplication part. After running the run_lsh.py, I got a file like en_head.clusters.parquet. I am not sure if I am right, but it seems to be the centroids of each cluster. How to obtain duplicates (the data I should remove from the dataset) based on the clusters file.

cq changed discussion status to closed
Together org

sorry for the late response here -- your are right, these contain the cluster ids, as well as the ids of documents for that specific cluster. To keep only one member of any cluster you can keep the document whose id (int) corresponds to the cluster id.

Sign up or log in to comment