Picture books: Automated story generator

Date of Publication

2008

Document Type

Bachelor's Thesis

Degree Name

Bachelor of Science in Computer Science

Subject Categories

Computer Sciences

College

College of Computer Studies

Department/Unit

Computer Science

Thesis Adviser

Ethel C. Ong

Defense Panel Member


Charibeth Ko
Nathalie Rose Lim

Abstract/Summary

Picture Books us a story-generation system that creates stories based on a given picture. The system creates an abstract story representation of the theme that matches the given picture and uses NLG techniques to generate the sentences of the story. The picture consists of characters, objects and a background selected from a pre-denied set of stickers and background images that enable the system to determine the essential elements of a story namely the characters, objects and setting. The generated stories are for children of age four to six.

Two manual evaluations were done to validate the grammar and content of the generated stories. The grammar aspect of the generated stories was given a score of 91.5% (3.66 out of 4 points) by the linguist mainly because of lack of transitional devices and possessive pronouns, while the content aspect of the generated stories was given a score of 88% (3.52 out of 4 points) by the child educators due to their opinion that stories with dialogues are more attractive to children rather than stories with simple declarative sentences. Two automated evaluation were also performed by comparing the generated stories against the reference stories validated by the linguist. Test results showed that the system has a low word error rate of 9.71% and a sentence error rate of 41.96% due mostly on the lack of transitional devices.

Abstract Format

html

Language

English

Format

Print

Accession Number

TU14609

Shelf Location

Archives, The Learning Commons, 12F, Henry Sy Sr. Hall

This document is currently not available here.

Share

COinS