Sparkly for blocking

[joemerfirearmland]

Hi, I am fairly new to entity resolution. It was mentioned in the paper that Sparkly was used for the blocking method, but in the code retriv.SparseRetriever was used, which is a basic BM25 implementation. Am I missing something?

[tshu-w]

@joemerfirearmland Thank you for your question. In fact, Sparkly uses TF/IDF similarity (BM25 in their paper) measurement for blocking operations (optimized for large data). When dealing with small-scale experimental data, for simplicity, I used retriv.SparseRetriever, which is a basic BM25 implementation.