Metadata priority & how does metadata lookup work?

[MelBr]

Hi all,

I know there's "cover" priority and you're able to rank covers based on the source they're coming from so that if you do auto-metadata download, it will prioritize it based on that table.

However, how does Calibre choose metadata that it retrieves from multiple sources? How does it pick one over the other? Could someone please explain it to me?

Does Calibre merge data from several sources? When you're presented with choices to pick metadata from, why is there no source label on these choices that shows you where it came from?

As I've mentioned earlier, I've had pretty bad experience with auto metadata lookups and I've ended up with a large number of books that had foreign or just messed up metadata.

Currently, I set my only source of metadata to be Goodreads and perform mass-lookup there and then I check off Amazon/FF/B&N/Google and perform manual, item-by-item, lookup on books that GR failed on. (GR is probably the best source of general fiction matadata right now)

I wish I knew of a way to prioritize it however... something like: look-up on GR, if it fails, lookup on Amazon, if it fails, try B&N etc...


Thanks for any insight!

[BetterRed]

MelBR,

are you 'priming' the metadata identifier with the ISBN before looking for Metadata.
There's an Extract ISBN plugin that will do that automagically.

kiwidude wrote some useful insights on this recently - my recollection is that there's no silver bullet... sorry I can't recall where he wrote about it, you could use search to browse his recent posts in the Calibre forum.

BR

[MelBr]

BR,

Yes, I do prime them using that excellent plug. Sometimes the ISBN extracted is wrong (not sure why), sometimes the metadata list contains a bunch of entries that are foreign versions of that book and sometimes I don't get a number out of it at all (djvu, scanned pdf etc).

I'll try to find kiwidude's post on metadata! Thanks for a pointer. I'd really like to learn how Calibre picks entries during mass-download of meta. I have a whole bunch of books and doing it all manually just takes way, way too long. I don't mind doing it for few books but doing it for 30-40% of them is just a giant pain. I just don't want to end up with a bunch of mislabeled books. I'd prefer my current way of finding things (search of filenames) to having books completely mislabeled as something else and not being able to find them anymore.

PS: I've PMed kiwidude few days ago asking about how to write a plug for a site (academic books one) and he gave me a few pointers! I'll try to write my own plug in next few weeks. Don't want to bother him again

[DoctorOhh]

Originally Posted by MelBr

I'll try to find kiwidude's post on metadata! Thanks for a pointer. I'd really like to learn how Calibre picks entries during mass-download of meta. I have a whole bunch of books and doing it all manually just takes way, way too long. I don't mind doing it for few books but doing it for 30-40% of them is just a giant pain.

Kiwidude's post essentially says to do it one book at a time and frequently you may have to do it a second time to get the GoodReads plugin to download data. The first time uses other plugins to prime the identifier so GoodReads has a better chance the second time around. He states that there is no substitute for actually reviewing the info.

[MelBr]

Originally Posted by DoctorOhh

Kiwidude's post essentially says to do it one book at a time and frequently you may have to do it a second time to get the GoodReads plugin to download data. The first time uses other plugins to prime the identifier so GoodReads has a better chance the second time around. He states that there is no substitute for actually reviewing the info.

OK, so I guess there's no silver bullet right now.

Priming with ISBN does help (when the number extracted is right) but GR is by far the most reliable way to update popular fiction books even without ISBN present. I really haven't had many false positives with it.

As I start importing my science & non-fiction books, GR will be useless and I'l have to deal with "meh" sources like Amazon and garbage sources like Google.