Casey Gollan â„ liked @thisisaaronland@orthis.social 2 years ago t thisisaaronland @thisisaaronland@orthis.social · 2y đ did i subconsciously know / when I made this drawing three months ago / that it would end up in a blog post about using vector embeddings to facilitate address de-duplication / probably â https://whosonfirst.org/blog/2024/08/16/dedupe/Who's On FirstDe-duplicating Who's On First venues with vector embeddingsUsing four different Whoâs On First venue repositories for testing, I have been able to first deprecate about 45,000 duplicate records and then, second, derive over 100,000 concordances with Overture Data place records, 8,000 concordances with All The Places venues and another 500 concordances with ILMS museum records. There are almost certainly still bugs, or at least âgotchasâ, but importantly the work so far passes the âbetter than yesterdayâ test.
t thisisaaronland @thisisaaronland@orthis.social · 2y đ did i subconsciously know / when I made this drawing three months ago / that it would end up in a blog post about using vector embeddings to facilitate address de-duplication / probably â https://whosonfirst.org/blog/2024/08/16/dedupe/Who's On FirstDe-duplicating Who's On First venues with vector embeddingsUsing four different Whoâs On First venue repositories for testing, I have been able to first deprecate about 45,000 duplicate records and then, second, derive over 100,000 concordances with Overture Data place records, 8,000 concordances with All The Places venues and another 500 concordances with ILMS museum records. There are almost certainly still bugs, or at least âgotchasâ, but importantly the work so far passes the âbetter than yesterdayâ test.