rzo1 commented on PR #635: URL: https://github.com/apache/opennlp/pull/635#issuecomment-2218458211
I am going to close this PR for now. Given the fact, that the f-measure on `main` was created as a baseline without using an actual cache, this PR should create the same results with caching. However, this is not the case, which implies, that the changes are not valid. However, there are still some open questions to better understand what actually happens here and how we can implement a more efficient way of caching something without changing the f-measure in the eval test. The following questions remain: - Why did we originally choose to use `==` instead of `Arrays.equals(...)` for caching the features. - Why does using `Arrays.equals(...)` only impact the eval tests for Dutch. Side-note: Played around with `IdentityHashMap` to get a similar behaviour as on `main`. This will make the eval tests pass (as expected) because the cache numbers are identical, but doesn't make sense to impl it like that (because no gain) -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
