Software as an editor – will we soon find new bestsellers for publishers?

Osnabrück. Today is World Book Day. In Germany, this day has established itself as a nationwide reading festival. But what do Germans actually prefer to read? An answer to this question promises a software for predicting book success.

This was probably the most expensive mistake in literary history: about a dozen British publishing houses had rejected the manuscript for the first Harry Potter volume by Scottish author J.K. Rowling before Bloomsbury finally took pity.

The no-sayers lost billions in profits. The seven books on the sorcerer’s apprentice of the 21st century have sold more than 450 million copies to date. The authoress and former welfare recipient Rowling accumulated an estimated 589 million euros in assets.

Happy coincidence

Rowling owes the fact that her first book ever found a publisher to a happy coincidence: the head of Bloomsbury had taken home the first chapter of her manuscript and had given it to his then eight-year-old daughter to read. She persuaded her father to give “Harry Potter” a chance.

According to the Börsenverein des Deutschen Buchhandels, around three million manuscripts are also received annually by German publishers. At the same time, more and more proofreading jobs are being saved. It is inevitable that many of our manuscripts are only superficially checked or in the worst case even sent back unread.

“Lisa” doesn’t judge

The cultural and media scientist Gesa Schöning and the programmer Ralf Winkler have now developed software to prevent bestselling manuscripts from being included. Its program, called “Qualifiction”, is based on data from around 20,000 bestsellers and non-bestsellers from past years and is constantly learning new things. The data are subjected to various analyses that are intended to help publishers, editors and authors make decisions.

“Qualification” consists of two parts: First there is the “Lisa” feature, which, according to Winkler in an interview with our editorial staff, is “purely descriptive and does not carry out any evaluation”. “Lisa” screens and analyzes new texts within the shortest time according to different aspects. In the context of the style analysis, for example, users learn how complex or simple the language of the work is.

The sentiment analysis, which identifies different moods, provides information about the tension curve, among other things. As Winkler explains, the sentimentvalue of almost all fiction works is very often in the slightly negative range. “This is because there should always be a basic tension in every good novel to avoid boredom,” says the expert. The value ranges between the extremes -1 for “devastated” and +1 for “sky-high cheering”. Descriptions of interpersonal and non-verbal communication should also be well received by the reader.

Variety of themes of a work

The software also focuses on the variety of topics in a work: “It is very bad, for example, if the author tries to deal with very many topics simultaneously in a book, and all of them are equally important,” says Winkler.

The second part of the software is called “Bestseller -DNA and determines the probability of success of a manuscript within a few seconds. To do this, he scans it for patterns and codes that make up a customer. If you want to know more about it, you will also receive an indication of the recommended circulation.

A similar procedure has already been successfully tested at Stanford University: In their book “The Bestseller Code”, editor Jodie Archer and English professor Matthew Jockers describe how the analysis program, fed with 5000 novels, calculates the probability that a work will be on the New York Times bestseller list. The hit rate is said to have been 80 percent.

Winkler does not share concerns that the software could help publishers publish only mainstream titles and that creative authors and titles might fall by the wayside. On the contrary: the IT expert believes that the use of “qualification” even increases the chances that “works that are off the mainstream or by lesser-known authors will even be viewed in publishing houses”.

By quickly screening a text with “Lisa”, editors could, for example, become aware of an extraordinary thematic cocktail that they would otherwise probably have missed. “Novelty can also be a success criterion,” says Winkler.

A lot of thrillers

In the meantime there are many thrillers that nobody wants to read anymore, because they are basically copies. “If there are thousands of comparative works from one work, all of which have not been successful, the machine learns that, of course, and draws its conclusions from them.

Qualifiction” is causing a sensation in the book industry: “We are currently selling the software to the first publishers,” says Winkler. Its further development is also supported by an EU programme and the Federal Ministry of Economics.

When Joanne K. Rowling submitted the manuscript for her first crime novel “Der Ruf des Kuckucks” under a pseudonym in 2015, several publishers reacted with cancellations. Among them were some who had already rejected the first Harry Potter volume almost two decades earlier. This might not have happened with the “Bestseller DNA” software.