Machine read CA; a path for those less enthralled by IT?

Tags: #<Tag:0x00007f050bb50290>


While going the whole hog and tracking down every last relationship for a release is a very good way to get to know your music collection - and probably does make it sound better to you - it does take time.

If a MB contributor is good at making scans of cover art, they can possibly make a valuable contribution, with less tech-headaches, by producing CA ( cover art) scans for releases without them. Even if they just add the minimum required metadata for a release as long as they have contributed the tracklist CA, booklet CA, front CA and a DiscID then all that is available for others yo work from, or even to be machine read by AI in a decade or three.

If someone found this mode of contributing fulfilling they could even come to some agreement with the local second-hand shop whereby they rent a box of CDs for a week, scan and enter the basics and then return and exchange them for fresh CDs.

A low cost, lower tech, more visual way to work towards high quality data.


In the end cover art is the only way to achieve high data quality, because it is the most original source we’ll ever have for most releases. What you suggest is already possible of course, but maybe MusicBrainz could make it a clearer option for new editors or even ‘advertise’ it?

We should get a few records store owners hooked up on MusicBrainz. :slight_smile:


I know that Open Food Facts uses machine reading of images as a source of their information, but most of their information input is also a lot more freeform (though structured, akin to our event tracklists), so they’re basically just “reading” e.g., the ingredients list and pasting that as-is into their “ingredients” field.

Not sure if maybe it’d be worth it to contact them about their experiences with this kind of technology.