Skip to content

OPLIN 4Cast #251: Automating readers’ advisory

Posted in 4cast’ advisory has long been an important component of the librarian’s job. People coming to libraries expect the librarian to be able to recommend a next book to read, and while much of successful readers’ advisory depends on skillfully leading the reader through a conversation that will reveal their interests, it’s also very useful if the librarian has access to databases that link similar books to help her/him make some targeted recommendations. Lately, several services on the web are trying to improve their book databases to the point where they could be more effective, self-contained tools for helping people discover new books to read, in effect automating the readers’ advisory process. Will they be successful?

  • The evolution of data products (O-Reilly Radar/Mike Loukides)  “Discovery is the key to building great data products, as opposed to products that are merely good. The problem with recommendation is that it’s all about recommending something that the user will like, whether that’s a news article, a song, or an app. But simply ‘liking’ something is the wrong criterion. […] I need software to tell me about things that are entirely new, ideally something I didn’t know I’d like or might have thought I wouldn’t like. That’s where discovery takes over.”
  • Using 20 billion data points, Goodreads will recommend your next book (ReadWriteWeb/John Paul Titlow)  “When most people hear ‘the Netflix of book recommendations’ they tend to think of another Internet giant known for its powerful recommendation engine: Amazon. Goodreads says it can provide better book recommendations than Amazon can because it has more data about what people actually like and dislike, as opposed to just purchases, browsing history and ratings.”
  • From commentary to conversation: the evolution of social reading (Publishing Perspectives/Matteo Berlucchi)  “Imagine therefore a Wikipedia style service which allows any reader to create a topic, add a collection of relevant books to that topic and let everyone else add more relevant books while also ranking the most interesting ones in order of preference. This ‘reader-generated’ topic system could grow to offer multiple ways to discover books by simply letting people browse these ‘virtual tables.’”
  • Hooked on context (Interview with Valla Vakili/Jenn Webb)  “We go through and create a graph for all of the little things inside of the books — the things that lead you off to new places — and then we show you all of the books that share those same elements. Once you’ve read the book, you can decide that you just want to go get the music, or you can decide to go get the music and then discover other books that have similar kinds of music in them. It’s two types of discovery: The first takes you deeper into the world of the thing you’re already in — places and things and such — and the second leads you toward books like the one you’re reading based on the objects that we’re graphing.”

Book data fact:
Any good book recommendation web service will depend on massive amounts of data about massive numbers of books. More than 300,000 books are published each year, and self-published ebooks will quickly drive that number even higher.