Slashdot Mirror
App Gives You Free Ebooks of Your Paperbacks When You Take a "Shelfie"
Peter Hudson writes Alan Henry writes on LifeHacker: "Paper books are awesome, but sometimes there's no beating the portability of an ebook on your phone or tablet. If you have a physical book you'd love to read on the go, BitLit may be able to get you an ebook version for free—all you need to do is take a photo of your book case: a 'shelfie.'" CNET notes that it's not quite as useful as it sounds: "As you might expect from a startup in the e-book space, BitLit currently offers a very limited selection -- only about 75,000 books, so the likelihood of a match is pretty slim. Browsing the library, I recognized very few mainstream authors."
I'm the app's developer... If anybody has any questions, I'm happy to answer them. I should probably get these out of the way first because this is /. 1. Yes it's possible to cheat the computer vision algorithms with the following: A dry erase market and an acetate sheet; a high quality color printer; a dark pencil that you really grind into the copyright page. 2. No you can't get free ebooks by taking shelfies in libraries or bookstores. Not unless you're also willing to write your name into said paper books. 3. If you really want an ebook there are probably easier ways to download one without getting out of your chair. 4. There are 234 publishers on board (as of this morning). They offer about 80,000 titles. 5. About half of those 80K titles are public domain and we give them to you without requiring that you sign the copyright page. 6. Of the half of the 80K titles that aren't public domain about 30% are free and 70% are paid. 7. Of the 70% (of the 40K copyrighted titles) the average discount you'll get is 86% off the digital list price... so when we say "highly discounted" we think that's fair. 8. Of the 234 publishers on board, I can count on two mittens the number who require us to use DRM. 9. The LifeHacker and CNet stories totally blew our servers up and my co-founder and I have been scrambling to get more AWS instances up and the back end distributed across them so that OCR can run faster... but as it stands the shelfie queue is about 13,000 deep... so the 15 minute processing time that it says in the app is wrong by about 6 hours. Summary: Yes you can cheat, but it's probably easier to cheat in other ways. Lots of titles, the majority are free as in beer, and almost all are free as in speech.
Great questions... sorry the actually method wasn't super clear in my first post. Here's how it works: you take the shelfie and the algorithms will identify every book on your shelf regardless of whether or not it's available... so even if you don't have any eligible books, you still get a tool that'll quickly and painlessly inventory your entire library. Once you've taken the shelfie and all the books have been ID'd we'll send you an email with download links to any books you own for which we have public domain ebooks. For any paper books you have which are not yet in the public domain, but which are from the 234 publishers that we've signed deals with, these books will show up in a list of "eligible" titles in the app. To claim an eligible title you have to take a picture of your name written onto the book's copyright page. As we sign up more publishers we will push a notification to the app if you have a title from a new publisher that's part of your shelfie.
The app is currently live in Canada, US, UK, Ireland, Australia, New Zealand, and South Africa... publishers often sell rights into different markets, so publisher A might own the digital rights to a book in the UK and publisher B in France. We get territory information from publishers in a format called ONIX which makes it easy for us to offer the books that the publishers we've signed with have the rights to offer in the territories that can offer them. So that's one of the reasons we've restricted the area of distribution. The other reason is the language model we use for the OCR in the shelfie. Since we're reading spines and since OCR is never 100% accurate we've got to apply an auto-correct model against a language database. We've used the English language "books in print" database from Bowker for this. There's no reason we can't expand in more time... but we're only a team of 8 of which only 4 are technical. So it's more of a bandwidth issue than anything.
We don't have access to every cover, but we do have a surprisingly good database of book covers. I think we have about 15 million or so... but we actually only need good cover art for the books from the publishers we've signed with (and they're happy, of course, to give it to us)... for all the other books that we'd identify in a shelfie we're actually using a deep neural network to read the letters off the spine and match that to a database of books in print.
The legality of it depends on the copyright laws of the country you live in... I'll let the ethicist from NYTimes magazine answer the matter of whether or not you have the moral authority to pirate an ebook if you own the paperback.
Sorry "books in print" is actually a product/brand name for Bowker's database of book metadata. It's a database of about 15 million records. Although it's expensive for us to license it, they do a lot of work hand cleaning up the records to ensure that spellings are consistent and duplicates are merged.
Because it's a the perfect combination of correct-me-if-you-dare-bait and neologism-portmanteau. That, and I couldn't wipe the shit eating grin off my face for days after I thought of it for the first time. Judge me accordingly.
This app demands access to your google account, and despite the fact that it claims only to want your "basic" profile information, I received complaints within minutes from people in my address book that I was sending them "spam" advertising this app.
Also, it uses a Web API to do it, so PrivacyGuard will not help you.