Intellectual Property
Patents
Two patents developed at ScriptBook by inventor Nadira Azermai — covering AI-based box office prediction from screenplays, and the automated generation of scripted narratives using machine learning. The research behind both directly informed DeepStory AI.
ScriptBook's technology was patented in both Europe and the United States, representing years of original research into how machine learning can understand, analyse, and generate screenplays — before large language models made the concept mainstream. Both patents are invented by Nadira Azermai.
EP3340069A1 · European Patent · Also registered in the United States
Automated Script Analysis & Box Office Prediction
What it covers
ScriptBook's core patent covers a computer-implemented system for analysing a screenplay using machine learning and natural language processing — and from that analysis alone, generating quantitative predictions about a film's commercial performance, target audience, MPAA rating, and critical reception.
The system processes a screenplay as a structured text document, extracting features related to story structure, character dynamics, dialogue patterns, genre markers, and narrative arc. These features are used to produce financial forecasts and audience analytics, with documented classification accuracy of up to 86% for box office success and failure.
This technology formed the operational core of ScriptBook's commercial platform — used by film producers, distributors, sales agents, and financiers across Europe and North America. The underlying technique for deriving static script vectors (used to represent a film's narrative fingerprint) is cited in the second patent below.
US20200334336A1 · United States Patent Application · Filed 2019 · Published 2020
Generation of Scripted Narratives
What it covers
This patent covers a novel machine learning method for automatically generating scripted narratives — screenplays, TV scripts, theatre plays, and game dialogue — that goes significantly beyond prior art by introducing what ScriptBook called "story-awareness."
The core innovation is representing a screenplay internally as a sequence of annotated sentences, where each sentence carries a paragraph type (scene header, action, dialogue, parenthetical) and character reference tokens linked to specific character identities. This architecture enables the model to maintain character consistency, narrative coherence, and structural integrity across long-form generated text — a problem that defeated all prior approaches to automated screenplay generation.
The patent further covers: character DNA profiling (per-character encodings that evolve dynamically across scenes); static biasing narrative encodings (steering generation toward a desired style or genre based on an existing script); positional encodings for controlling script length; synopsis-based generation; and a discriminator component that evaluates and rejects generated sentences below a quality threshold.
This research directly underpinned DeepStory AI — Nadira Azermai's proprietary story generation platform, built before the world knew what GPT was. The patent has since been cited by five subsequent patents, including filings by Oracle International Corporation.