OverDrive is preparing to launch Readtelligence, a suite of new features for ebook selection and curation developed using artificial intelligence (AI) and deep learning tools to analyze every title in the company’s inventory. Readtelligence was announced on August 5 during the “Your OverDrive Service of Tomorrow: Advancing the Science” presentation by OverDrive President and CEO Steve Potash at the company’s biennial Digipalooza conference, held virtually this year.

“Readtelligence is a series of routines and tools…that are the result of several years of R&D here at OverDrive Labs,” Potash said. Every title in OverDrive marketplace “is now being indexed [with] deep learning using AI technology, where every byte, every character, every word, image, sound, voice, style, format, cover image, is going through deep analytics. What we want to do is say, ‘I know we support MARC records and, of course, BISG and BISAC subject headings, grade leveling, genres, etc.’ But there’s been problems and challenges. I’ve heard from educators who say, ‘I started with a series of juvenile titles, and by chapter six, the characters are having sex.’ Or, ‘there is explicit material that I did not want to discover in this book.’ We have new tools. We are calling these tools Content Awareness Tools. This is the byproduct of [AI] crawling through [OverDrive’s ebook catalog] with a dynamic, growing list of attributes that are being reviewed and being scored.”

In addition to flagging unwanted content within an ebook, the tools can be used to highlight themes, or the most frequently mentioned locations or people within a title, for example. New metadata such as the average length of words, sentences, and chapters can be generated, as well as the estimated time it will take for an average reader to complete any title. An AI-generated “emotion curve” will offer librarians a visual glimpse of a title’s narrative contour. This granular data will enable new ways for librarians to search for similar content and curate unique collections. Although he was not conducting a live demo, Potash used book covers as an example, first showing the results for a search for ebooks with cats on the cover, and a second search for ebooks with purple covers that also have cats on them.

Using the virtual meeting’s live chat function, a few attendees expressed apprehension about Readtelligence’s Content Awareness Tools, noting that they could be used to block certain types of content from being considered for purchase. “Professionals better be on the same page about censorship, or ‘content awareness tools’ will enable us to make decisions we shouldn't make for our readers in the public library space,” one attendee commented. Another wrote “dictators would love this level of control over what their people read.” And another wrote “I can see content labels being useful because I know many readers who don't like violence or swearing, but these labels will need to be neutral because there's nothing wrong if a patron does like these things.” Yet another said that they would still be more likely to rely on tried-and-true metrics for choosing content, noting that “some of this may be useful, but as a selector I only have so much time for selection detail. If it has a long waiting list, it gets bought.”

Readtelligence will offer OverDrive’s customers the opportunity to use AI driven analytics tools similar to those already in use by publishers and companies such as Amazon. “I can tell you, there is demand for this,” Potash said, citing a recent job listing posted by Penguin Random House on Publisher’s Lunch, seeking a full-time AI and machine learning associate. “It’s not just OverDrive doing this. The booksellers, Amazon’s been doing it forever, and now you, the library, should have access to this kind of deep learning.”

Potash compared the at-a-glance data that Readtelligence will generate to the uniform nutrition labels on food packaging. “If the food industry 30 years ago could create an easy consumer guide to what’s inside the box, why can’t OverDrive take the insights from hundreds of billions of words and phrases, titles, images, and even the tone and emotion within a book using AI, and possibly report for you, the selector, the curator” granular information about new or unfamiliar titles, he said. “We’re going to let you know, objectively, what are the things you may want to be aware of…. We don’t set the policy, but we want to enable you to have this Readtelligence on every title, every supplier.”