HOMER: A Creative Story Generation System Student: Dimitrios N. Konstantinou Supervisor: Prof. Paul Mc Kevitt School of Computing and Intelligent Systems Faculty of Informatics University of Ulster, Magee
Objectives of HOMER To build a creative storytelling agent that generates: style-constrained stories stories with a point of view natural language output domain-independent stories
Literature Review Creativity Systems: Copycat Genesis Letter Spirit A Computational Model of Music A Computational Model of Poetry
Schank’s Theory of CD, Scripts, and Stories A Robotic Storyteller An objection to Schank’s Theory Scripts and Point of View Story Grammars
Comparison with other storytelling systems
The Rationale for HOMER Approximate the creative conceptual space in human narratives Transform the conceptual space Build domain-independence Develop an extendable creative agent Simulate author goals
Motivation Create associative clusters of variables (“archetypical modes”) Simulate high-level style decisions in story output Introduce mid- / low-level style decisions Simulate point of view Create lexical entries and use transformational procedures
The Parser Input frame Language Understander Inference Mechanism Style Specifier Frame Constructor
The Story-outline Constructor Frame story-outline Mode-based Hierarchies Mode-based imagery
The Natural Language Generator Story narrative outline Narrative Reasoner Text Planner Narrative History Revisor Ontology Surface Realizer
Conclusion HOMER: a creative storytelling agent that: takes as input a story fragment, style specifications and contextual clues simulates authorial creative goals generates narrative in the form of natural language output approximates human language output
Software Tools Analysis Parsing: Attribute Logic Engine (ALE 3.2) Lexical Knowledge Base (LKB) Rhetorical Structure Tool (RST) Natural Language Generation: Upper Model from KPML Systemic Unification Realization Grammar of English (SURGE) I-SAURUS, lexicon for near-synonyms
Storytelling Systems Structure-based Vs. Environment-based Automatic Novel Writer (1973) HOMER (2004) AESOPWORLD(1996) Larsen & Petersen (1999) Story elements: story-line, plot, setting, style, point of view Story Understanding Creativity
Project plan
Architecture of HOMER Input output Parser Thematic Memory Story-Outline Constructor Natural Language Generator