WaveAccess developed a recommendatory system for book service Get Your Book

2020: Creation of a recommendatory system

On February 19, 2020 WaveAccess announced creation for Get Your Book service of a recommendatory system based on Machine Learning (machine learning, ML). The solution creates the list of the recommended literary works on the basis of the interests of the reader.

The Get Your Book service offers users a monthly subscription to gift boxes: in them the buyer receives books, chocolate, tea and other trifles. Thus the user regularly fills up the book shelf, saving time and getting unusual experience. The choice of the book becomes a surprise: Get Your Book selects works, analyzing reviews of the leading critics, – only the published books and best-sellers get to sending.

To solve a sales increase problem at cost reduction and to increase loyalty of users, founders of the company decided to resort to the innovation approach: automate selection of books and personalize the offer, at the same time having saved criterion of "surprise".

Get Your Book addressed to WaveAccess – development of a recommendatory system based on machine learning which allows to select on the basis of the interests of the reader from extensive base of books top-25 works became the solution.

The owner of a subscription of Get Your Book fills out on the website the questionnaire in which specifies the genres interesting him, favourite works and authors. It is possible to specify one or several genres: art, classics, fantasy, self-development, non-fiction.

On the basis of these data a system creates the list of 25 recommended books, sorting them by the probability of coincidence to the interests of the reader. This list is available only to the staff of Get Your Book – the effect of surprise for the buyer does not vanish. At the same time existence of such recommendations repeatedly increases the probability to please the reader.

The algorithm of machine learning analyzes more than 8000 Russian-language books from open directories, including AST publisher and Hit-kniga shop. Missing data and reader's ratings "are tightened" from the website Good Reads on a unique identifier of the book of ISBN.

The developed system has high potential for development. She can be "taught":

• process series of books;
• consider not pleasant books;
• simplify a data entry format;
• bring together users with similar literary preferences;
• distribute the recommended books on the level of purchase prices that would allow Get Your Book to reduce cost value of sending.

The recommendatory systems serve as one of the most popular and monetized examples of use of solutions based on artificial intelligence and machine learning. At the expense of relevant recommendations to clients in the right place, the necessary time and via relevant channels they help business to increase conversion. At a recommendatory system for Get Your Book high potential, and we will be glad to offer the appearing opportunities for its development, comments the senior vice president for software development in WaveAccess Alexander Azarov