Google recently introduced an API to their book data. I want to add it as an Add-Books search.

The API has various features we already support, like links to Google Books, and ones we will support soon, like embedded pages. But it also represents a large, fast and fairly accurate collection of metadata about books—and possible alternative to Amazon and library sources.

Here's a URL for you to test it:
http://www.librarything.com/test_googleA...

The trick is, the data isn't very full. It has title, author (creator), date, ISBN, format, publisher and "subject."

The title is dicey. Subtitles are given as second titles, but sometimes the second title is just an alternate title or the same title again. Authors are given in first-last format, so LibraryThing has to guess what the last name is and what the first name is. Mark Twain can, of course, easily become "Twain, Mark" but what about "American Society of Psychologists"—is Psychologists the last name? ISBN is the only identifier, with no Library of Congress Numbers or OCLC numbers. Format is usually pages, but not always, and lacks dimension information that—although we're not using it now—we could use it from Amazon and libraries. Subject is sometimes one or more Library of Congress Subject Headings, with no hierarchy I've found, and sometimes BISC headings—it's completely useless.

In addition, the data has no "role" data, LCC classification or Dewey classification. We can't get LCSH from it. The source library is never named.

I plan to add it as a source. I think it will prove useful, particularly to users who want speed and are less interested in depth of data. It should be good for international, but I haven't tested it much there.

Message edited by its author, Sep 27, 2008, 11:59pm.

Did a short run (i.e. I entered one ISBN). The ISBN I looked up was 9062835996, which is the one of my second edition of Een kennismaking met de oude wereld from 1996.

And I was wondering about two things: what are the two identifiers I get back: VnwaAAAAIAAJ and UCAL:B3893129?

Also not that it actually gives back the data for the wrong edition: the first edition from 1983.

I think VnwaAAAAIAAJ is an internal Google number. UCAL:B3893129 is some University of California code. Apart from ISBNs, I haven't seen any "real" ones—LCCNs, etc.

Nobody else wants to comment on this? I'd love to get some feedback before we add this as a search option. My inclination is to make is on the default list for all users.

Keyword searching (by which I intended partial author and title) is awfully broad. For instance, I got The New Encyclopaedia Britannica and A History of English Philosophy in the results by searching for some famous ones. The actual books I intended seemed to be at the top.

I'm guessing it's their bug that Creator: University of Toronto Press, Victoria College (Toronto, Ont.) turns into three values because there are two commas.

Message edited by its author, Sep 28, 2008, 2:54am.

Probably the people most likely to post comments are the ones least likely to use a source that provides so little depth of information.

I took a look at it. It works OK for ISBN searches, and does a better job of coming up with results for random foreign archaeological monographs than any other single source is likely to. It isn't so good for author-title searches, though. It tends to spit out a long list of results that include a lot of books that probably *cite* the work being searched for. If it's widely cited, the actual work may not make it into the list of items shown (the first 25 results, I think?).

So I'd say it would be great for people who want to do automated imports of lists of ISBNs, but it may need some work before it's ready for other kinds of use.

If it had LCCNs or OCLCs then I could try to augment its results by searching on those identifiers at a "real" source—Google plus Library. As it is, I could probably come up with mapping from library-specific identifiers, like UCAL:XXXX, to the actual libraries, so when you clicked on a book it would go to the *library* for more information.

I think it could be a good source to add wishlists, and I'm going to poke someone who mainly adds books via ISBN about it.

> 6 I could probably come up with mapping from library-specific identifiers

You're probably right, but it wasn't obvious when I tried last night. I plugged the IDs into their OPACs and came up empty (no big surprise). I then found the record (by title) and looked at the details. In the case of Mirlyn, it appeared that the ID was of the digital artifact, not the original from which it was scanned. That is, it was a second "location" and directly mapped to the Hathi Trust URL. In the case of CATNYP, it wasn't anywhere to be found in the record.

It's hard to know what to do with it. It's so good in some ways. But it's just not there yet in others.

9: Then it matches perfectly with Amazon. ;)

Is Google Books data viewed as serious competition for Amazon data? I have read mixed accounts, and was wondering if it seems that it may develop into an alternative (to then free LibraryThing of some Amazon dependence. But, "out of the frying pan, into the fire", perhaps).

Would it make sense to pick this up again with an eye specifically toward keeping a collection synchronized with Google Books My Library?

So, the Google Books synch is interesting—something we've thought about—although it's separate from this. We could make something that took the pain out of synching with Google Books, although it would require adding books to Google that weren't exactly the right one. But I dunno. It's a lot of work, and Google has not been particularly giving to us for us to load millions of LT books into their system, which is an incipient competitor anyway.

What I want is a collection in LT that contains the thousands of public domain works I have bookmarked in GB, many of which I have downloaded as PDFs. (I don't bookmark preview or snippet, but that's just me.) Then I can call up all the classic works on X from one place. And I want this to update as I bookmark more.

I have no particular desire to update anything in GB from LT, but I could see that others might.

I could perhaps be retrained to only bookmark in LT, if there were a bookmarklet that made this almost as simple as "Add to my Library" in the GB reader. GB search within my library, even with religious tagging, isn't very good.

I'm willing to do part of this myself, though experience suggests that an experimental LT API is more work for you guys than an experimental LT PHP.

My experiences of 11 searches are as follows:

Eight satisfactory searches on ISBN:
http://books.google.com/books/feeds/volu...
http://books.google.com/books/feeds/volu...
http://books.google.com/books/feeds/volu...
http://books.google.com/books/feeds/volu...
http://books.google.com/books/feeds/volu...
http://books.google.com/books/feeds/volu...
http://books.google.com/books/feeds/volu...

One ISBN search yielding an incorrect publisher:
http://books.google.com/books/feeds/volu...

One satisfactory keyword search:
http://books.google.com/books/feeds/volu...

One keyword search not finding the book in first 25 hits:
http://books.google.com/books/feeds/volu...

I would be happy to use this source.

>> I would be happy to use this source.

me too! :D