Trajectory, a digital distributor and technology developer, has developed a proprietary algorithm platform it claims will take book recommendations to a new level of accuracy and utility. After scanning the text of a book, Trajectory claims its technology can deliver better keywords and book recommendations to its clients in the bookselling, library and school markets.

Trajectory is now offering commercial access to its Natural Language Processing Engine, a series of “deep learning algorithms,” said Trajectory CEO Jim Bryant, that will deliver a “big step forward in book discovery”—the ability to connect readers to desirable books they don’t already know about. The technology, he said, effectively maps and displays the personality of a book which can then be compared to other books.

Bryant said Trajectory makes money from the NLPE by charging publishers for keywords and "insight into their books." E-book retailers pay the company for book recommendations based on book information and "social vectors,' or book attributes they provide to Trajectory on a proprietary basis. Trajectory also distributes e-books to a global network of more than 300 retailers and librarians.

So far Trajectory has scanned about 30,000 titles, including the HarperCollins catalog (Trajectory acquired the assets of the now-defunct startup Small Demons, which included access to Harper’s e-book catalog) and about 1,000 public domain classics. Using networked computers, the Trajectory NLPE scans the text of thousands of books, turning each sentence and its components into language data patterns that can be analyzed and extracted. The NLPE charts more than 30 “attributes” in the text of a book, including length, chapter, pace, intensity, mood, word type, and reader age. Using this data the NLPE can identify other scanned books that share similar attributes and language patterns.

Rather than comparing a book buyer’s purchases for book recommendations—like Amazon and most other online retailers—Trajectory’s technology analyzes and reveals the linguistic structure inherent in the text of a book. The Trajectory NLPE can also capture the emotional content, or “flow of sentiment,” inherent in a book by tracking the occurrence of keywords. The NLPE, Bryant said, can extract a “fingerprint” or a “personality,” from the text that can be used to compare one book to another.

Bryant said the NLPE main products are keywords and book recommendations and the technology is on display right now in the Trajectory website. Under Discover Great Books, users can see book recommendations as well a breakdown of each book’s word cloud, attribute list and data visualizations. Under the Publisher category, visitors to the site can use the live Sentiment Graph, Keyword graph and NLPE database, and call up the graphical fingerprint of a particular book’s content and compare it to other books.

Bryant said the book recommendations from the NLPE are dynamic—recommendations will change and grow as more titles are scanned into the system. He also said he hopes to reach bulk processing agreements with self-publishing platforms like Smashwords in an effort to market NLPE keywords and book recommendations to self-published authors as well as booksellers and libraries.