What we have to work with, however, is simply the publisher name that we receive in the metadata, and that is generally the name that appeared on the title page of the book. There is nothing to link that name to an actual corporate entity (either the one owning the imprint at the time, or the one that may own it today). I think that our data is closer to “imprint” than it is to “publisher”, and that bringing the two together will require some external data. The ISBN prefixes may provide some help in that area. For example, code 0-02, which is generally listed as belonging to Macmillan, has appeared in metadata with these listed in the publisher field:

Macmillian
Collier Macmillan
Free Press
Maxwell Macmillan International
Maxwell Macmillan Canada
Collier Books
Macmillian Reference USA
Collier-Macmillan

If these are what you intend my “Imprint” then imprint is generally what we have in the data.

How are you managing “imprints” of publishers?

0-19 Oxford University Press 251368
0-16 U.S. G.P.O. 245442
0-521 Cambridge University Press 175340
0-415 Routledge 117731
0-13 Prentice-Hall 116635
0-471 Wiley 112967
0-06 Harper & Row 109797
0-07 McGraw-Hill 98202
0-312 St. Martin’s Press 91149
0-02 Macmillan 75461

And it’s interesting how many of the top ones are university presses.

Also, it looks like there are either typo’d ISBNs in the data (I’m sure there are some) or some publishers have shared ISBNs. For example, I find MacGibbon & Kee under a couple of different numbers assigned to others, as well as their own. Anyway, fascinating data here.