Caio L.

Commented on Using Zingg for KYC Entity Matching Without Positi...·Posted inHelp Zingg

Hey! Pleased to meet you too 🙂 We have strings for name, birth_date, gender, country and nationality (also a document_number but only for some of the entries)

Posted in Help Zingg·

Caio L.

Using Zingg for KYC Entity Matching Without Positive Training Matches: Best Approaches and Alternatives

Hello! Just found out about Zingg and I'm trying to understand if it could be useful for our problem. I have a KYC-related entity matching problem, but I don't really expect to have any matches (we're comparing our user database to a UN-sanctioned list of individuals). Therefore, I won't have any positive matches for training... Is it absolutely necessary to train custom models on Zingg? Or are there any standard models/configs that could apply to my use-case? I guess I will end up needing some synthetic data to evaluate any solution, and this data could be used to train a model, but I'd like to avoid that. My thinking is, if I don't expect to have any matches, maybe a simple approx. nearest neighbors approach like LSH is a better fit for my use-case... Any toughts here Sonal G.? Either way, Zingg is an inspiring open-source project, congrats on your work! 🥳

3Comments