email print share on facebook share on twitter share on google+


GoCritic! Industry: Analyse This


- At Karlovy Vary's industry programme, Nick Mastrini looks at how A.I. script analysis is affecting project development in the film industry on the example of the ScriptBook online service

GoCritic! Industry: Analyse This

‘Hard science. Better box office.’ The tagline of ScriptBook, an online service specialising in the analysis of scripts using artificial intelligence (AI), promises results in the film industry. But can its commercial remit be reconciled with the intricacies of human creativity?

The second day of the Eastern Promises industry programme at the 53rd Karlovy Vary International Film Festival opened with a presentation of ScriptBook led by data scientist Michiel Ruelens and the company's CEO Nadira Azermai. She began by noting that film distributors have ‘a lot of risk’ to consider in the subjective process of greenlighting potential screenplays. Using a dataset comprised of 6,500 scripts and 15,000 films’ meta-data, the service aims to reduce this risk by identifying the scripts with the most box office potential.

(The article continues below - Commercial information)

The situation calls to mind Michael Lewis’ 2003 book Moneyball, which depicts the tension between traditional baseball scouts and a new data-driven approach to the sport’s recruitment methods. Azermai describes the ‘unconscious bias’ of studios in their script selection, highlighting another stubborn status quo being challenged with the use of technology. With ScriptBook, she notes, there is "never an input from humans".

This was the crux of the initial backlash against the service, however, after it first came into the public eye when a partnership with script-survey website The Black List was reversed in April 2017. Anomalisa producer Keith Calder responded to that news by emphatically rejecting ScriptBook as "snake oil garbage masquerading as an 'objective' tool. Writers and execs, please give this a hard pass."

In response to this criticism of false objectivity, ScriptBook may thrive best if it markets its function as a comparative rather than definitive tool. The service’s use of machine learning accumulates a network of screenwriting data, with every detail assessed, categorised and described as ‘Script DNA’. Features include a graph showing a screenplay’s pacing, based on the distribution of dialogue and action, and a predicted MPAA rating, with reference to the scenes that censors may flag. These are objective metrics that can help to make a subjective decision.

In the 'Character Sentiment’ feature of ScriptBook, percentages are applied to the predominant emotions that a character’s descriptions and dialogue suggest. Such numbers can seem trivial, even inaccurate when compared with human judgement, but they provide a compelling reference point for writers to assess their use of language. While a writer will know their script intimately, they will be equally aware that another editorial perspective — whether human or digital — may be beneficial to the creative process.

Azermai argues that this provides financial benefits, as there is "value in preventing the production of certain projects". ScriptBook’s validation study with Sony Pictures notes a near 50/50 split in the studio’s box office hits and misses; when these screenplays were simulated through the ScriptBook platform, 22 of the 30 box office failures would not have been greenlit. But as in the case study of Passengers, although the box office forecast had a good degree of accuracy, the tool could not anticipate the poor critical reception of the film.

In fact, Passengers was a film fostered by The Black List, and now touted by ScriptBook as a sign of the service’s quality. Ultimately, the purpose of this script analysis is not to parse the subtleties of the screenplay, but to compare its content to the known quantities of previously released films. This has the benefit of signalling trends to writers and studios, and guiding their creativity towards originality.

By revealing the bias inherent in produced films, this technology can provide a shortcut to greater diversity in the industry. The tool inherits the progress made by the Bechdel Test and builds on it, analysing potential inequality word by word and detailing the distribution of dialogue between men and women across scenes and screenplays. ScriptBook may currently promise ‘better box office’ at the core of its service, but better representation should be its mission — and the technology has the potential to deliver it.

This article was written as part of GoCritic! training programme.

Did you enjoy reading this article? Please subscribe to our newsletter to receive more stories like this directly in your inbox.