audiobookshelf
audiobookshelf copied to clipboard
advplyr
[Enhancement]: Fuzzy Matching
Describe the feature/enhancement
Adding a new library to your set up, and I'm finding that the match patterns are quite a lot stricter than I think they have to be.
e.g.
Dots after abbreviated middle names:
- James SA Corey + Expanse book title no match
- James S.A. Corey + Expanse book title match
Book Title Match, author different
- Benedict Cumberbatch Reads Sherlock Holmes' Rediscovered Railway Stories + Arthur Conan Doyle no match
- Benedict Cumberbatch Reads Sherlock Holmes' Rediscovered Railway Stories + John Taylor (the actual author!) match.
I get the latter, but it's a very long name and seems like there should at least return some results here.
The Plex audiobooks metadata agent returns with a percentage match based on several criteria. If you're worried about your auto-matching breaking with fuzzy matching, why not just have a threshold?
@Albuca, I think the Audnexus Plex metadata agent does a pretty good job of polling the Audible API - I think it's less strict than your search. Don't know if you're familiar with Python, but it's a pretty simple setup with just a few files. Think they use some sort of scoring system to evaluate the returned results.
@jt196 Audnexus has been being used for all audible searches for a few versions now
@advplyr, great stuff. Did double check the test searches above and neither returned results. Guess the Sherlock one isn't the biggest deal, but James SA Corey. Is there a way to dial down the fuzziness? Even if it didn't auto match below a certain threshold that would be useful in the manual search to have less accurate results returned.
This was implemented pretty well for the next release. PR https://github.com/advplyr/audiobookshelf/pull/2099