Developing a Creative Story Teller Agent

Primary Faculty Mentor’s Name

Rania Hodhod

Proposal Track

Student

Session Format

Paper Presentation

Abstract

This paper focuses on developing a creative storytelling agent that makes use of commonsense knowledge bases to create new, interesting stories from simple story templates. The storytelling agent makes use of a commonsense knowledge base known as “ConceptNet”—an artificial intelligence project started in 1999 at the Massachusetts Institute of Technology-that is built from nodes representing concepts, in the form of words or short phrases of natural language, and labeled relationships between them. These are the kinds of things computers need to know to search for information better, answer questions, and understand people's goals. This paper describes the process in which the storyteller agent is able to logically place new words (retrieved from ConceptNet) into a story template. For example, if the template were to give the agent the phrase: “a boy name”, it can query ConceptNet and retrieve “Bob” or “Jack” on the basis of those being described as boys names. Similarly, the template can query “a girls name” or a “unisexual name” and return results that ConceptNet finds applicable. The result of these queries is a story that can feature many different elements from a previous one, yet retain the original core tale. One of the challenges that arose encountered while implementing this storyteller agent was finding a suitable “engine” or foundation in which to facilitate the process. Although there exist some off shelf storytelling engines, such as RenPy that allows animation and music in addition to the story, but ultimately was not suitable for the use of ConceptNet. This raised the need to develop our own storytelling engine—which is referred to as AstroStory. AstroStory replicates RenPy interface and design, but was more flexible from the backend in a way that allows more intimate interaction with ConceptNet. The AstroStory engine was implemented using the integrated developer environment (IDE): Xcode— an IDE exclusive to Macintosh systems. The interface and the layout for the storyteller were then tied all together to produce a working prototype—one that could follow a template given and create multiple stories based on it. A longer term goal with this project is to manifest this agent on a variety of platforms—but this requires some more optimization of ConceptNet and expansion of the knowledge base where the agent is able to include whole phrases instead of singular words.

Keywords

ConceptNet, Commonsense knowledge base, AstroStory engine

Location

Room 2908

Presentation Year

2014

Start Date

11-15-2014 1:45 PM

End Date

11-15-2014 2:45 PM

Publication Type and Release Option

Presentation (Open Access)

This document is currently not available here.

Share

COinS
 
Nov 15th, 1:45 PM Nov 15th, 2:45 PM

Developing a Creative Story Teller Agent

Room 2908

This paper focuses on developing a creative storytelling agent that makes use of commonsense knowledge bases to create new, interesting stories from simple story templates. The storytelling agent makes use of a commonsense knowledge base known as “ConceptNet”—an artificial intelligence project started in 1999 at the Massachusetts Institute of Technology-that is built from nodes representing concepts, in the form of words or short phrases of natural language, and labeled relationships between them. These are the kinds of things computers need to know to search for information better, answer questions, and understand people's goals. This paper describes the process in which the storyteller agent is able to logically place new words (retrieved from ConceptNet) into a story template. For example, if the template were to give the agent the phrase: “a boy name”, it can query ConceptNet and retrieve “Bob” or “Jack” on the basis of those being described as boys names. Similarly, the template can query “a girls name” or a “unisexual name” and return results that ConceptNet finds applicable. The result of these queries is a story that can feature many different elements from a previous one, yet retain the original core tale. One of the challenges that arose encountered while implementing this storyteller agent was finding a suitable “engine” or foundation in which to facilitate the process. Although there exist some off shelf storytelling engines, such as RenPy that allows animation and music in addition to the story, but ultimately was not suitable for the use of ConceptNet. This raised the need to develop our own storytelling engine—which is referred to as AstroStory. AstroStory replicates RenPy interface and design, but was more flexible from the backend in a way that allows more intimate interaction with ConceptNet. The AstroStory engine was implemented using the integrated developer environment (IDE): Xcode— an IDE exclusive to Macintosh systems. The interface and the layout for the storyteller were then tied all together to produce a working prototype—one that could follow a template given and create multiple stories based on it. A longer term goal with this project is to manifest this agent on a variety of platforms—but this requires some more optimization of ConceptNet and expansion of the knowledge base where the agent is able to include whole phrases instead of singular words.