I’m all for AI and chatbots – I even designed one! -, but every time I hear ‘use artificial intelligence for book discovery’, I’m sceptical. I didn’t see anything worth looking at since Oyster was closed down, and when I heard that Google has launched Talk to Books, I wasn’t sure what to think. I’m a big fan of everything Google, but most of their new features get shut down within a few months. (Remember Google Buzz?) Anyway, I have tried Google’s free to use AI features a few months back (Google Vision and Natural Language Processing), and I was hoping that Google will soon use these tools to offer better discoverability to the books in their catalogue. Maybe even increase sales?
If you are not familiar with Google Books: it is Google’s enormous project, scanning books from all over the world and making the content searchable. The project has not always been everyone’s favourite, but they have recently overcome some copyright challenges, and are working continuously on making information freely available for everyone. It is not only public domain books and legal documents that form Google Books’ catalogue: they have access to the text of all books distributed through Google Play Books as well. Publishers and authors signing up to Google Play Books agree to their books being part of the Google Books project. It doesn’t mean that the whole book will be available on Google Books. It will, however, make the text searchable.
Why the long lead in?
Because it is exactly the gargantuan Google Books database that is used as a basis of the new project, Talk to Books. (The other new lab project is called Semantris, and it is clearly ruining the effectiveness of my workday. It is a word association game: try it, and you will no longer need Tetris. Hey, don’t blame me!)
How does it work?
In Talk to Books, when you type in a question or a statement, the model looks at every sentence in over 100,000 books to find the responses that would most likely come next in a conversation. The response sentence is shown in bold, along with some of the text that appeared next to the sentence for context. (Semantic Experiences)
To put it simple: Talk to Books uses natural language processing and the full text of the books in its database to recommend you a book that would answer your question. It kind of works like an omnipotent librarian. If Talk to Books ever leaves the experimental stage, there is no more “ask Google”.
As the project is on trial, it is not (yet) using Google Books’ full database: only 100,000 books are searchable. So far, it seems to be working effectively for non-fiction: it gives you relevant results immediately. Please note that it prefers full sentences over simple keywords.
Wondering where to buy the book? Just click: it takes you directly to Google Play Books to make the purchase. If the book is not available as ebook, Google Books offers links to brick-and-mortar stores where the book is available.
What about fiction? Can Talk to Books give relevant recommendations from fiction books? I tried testing it with one of my guilty pleasures: women sleuth fiction.
Is it the way I ask? Or Talk to Google really cannot recommend me a single book I would like to read? Its database certainly contains fiction books (they sporadically come up in the search results), but the recommendation engine prefers non-fiction.
You should definitely head to the Talk to Books website and give it a go: it is fun and funny. It give us a glimpse into the future of what libraries and bookstores might be capable of.
But when it comes to answering a question, I’d still prefer a real life librarian.