Natural Language, Semantic Analysis, and Interactive Fiction (2006) [pdf] (worrydream.com)

Graham Nelson’s 2006 paper recounts the Inform 7 project—a deliberate experiment to let authors write interactive fiction (IF) in natural language by using a controlled subset of English as the programming syntax. Rather than traditional code, Inform 7’s “source text” reads like narrative; the environment is presented as a single “book” with facing pages (left: author text, right: game consequences), automatic project management (no filenames), inclusion-by-author-name (e.g., “Include the Automatic Door Rules by Emily Short”), and a rule-oriented architecture that replaces the usual object-oriented IF model. Nelson describes practical design choices (literate-programming style sections, automatic indexing, real-time word-wrapping) alongside the project’s UI and workflow choices born from testing and collaboration. The paper argues that semantic theory should sit at the heart of IF systems and draws on conceptual semantics, predicate logic and model theory to explain why natural language is especially suited to describing stories and situations. For the AI/ML community this work is significant because it reframes programmatic authoring as a form of narration and shows how formal semantic tools can guide the representation of events, states and rules in interactive worlds. Implications include richer natural-language programming, tighter integration of semantics into content generation and story modelling, and a challenge to entrenched object/inheritance paradigms—suggesting new research directions in narrative understanding, knowledge representation, and human-friendly programming interfaces.