openlibraryenvironment/gokb

Title mismatch Apollo

Closed this issue · 8 comments

Given the TitleInstance https://gokb.org/gokb/resource/show/org.gokb.cred.BookInstance%3A6241061#titledetails.
In fact, it should be two title instances given the same name because it occurs in https://gokb.org/gokb/resource/show/org.gokb.cred.Package%3A11751449 and in https://gokb.org/gokb/resource/show/org.gokb.cred.Package%3A6243109. In former, it is a film title and in latter, it is an eBook. We are thus talking about two different title instances what has the consequence that film titles get ISBNs.

No. Items from KBART can only be separated via their publication_type, namely serial or monograph. Those values directly correlate with a specific Item type in the GOKb (BookInstance/JournalInstance). The value of medium can then be used to further differentiate the titles, but will not be used for matching of future imports. So unless the titles contain differing global identifiers (so both would need a different doi in effect), I am afraid your use case cannot be reflected properly by the KBART standard and you should therefore refrain from misusing it in this manner.

The matter of this ticket is that two independent records are reflected into one.

This is because only one of the titles has any reliable identifiers, and both are classified as a BookInstance. Titles without proper identifiers cannot be matched properly, and are therefore prone to cause such errors. The safest solution would be to refuse any titles without proper (globally recognized) identifiers, but that would have to be discussed further.

Alright, should thus become subject of discussion.

Given the fact that countless examples of matching books & films titles exist, we need to make this work.

@hornmo : What is the minimum criteria to make two titles distinguishable in this case?

My assumption: if there is a duplicate title, check the medium type.

  • If the medium types are different, check identifiers.
    • If identifiers are different: OK --> create both items independently
    • If there is a duplicate identifiers: --> create a Review Request
  • If the medium types are the same, we can not defer a case descibed here. So the procedure must be kept as is.

Yes, that sounds correct. I was going to completely ignore other types of titles in the identifier lookup, but creating a review request could probably be helpful for such cases.

@hornmo Is this fixed within the two PRs mentioned above? If so, please close this issue.

Those PRs fixed this issue.