Extracting conceptual relations from children’s stories

College

College of Computer Studies

Department/Unit

Software Technology

Document Type

Article

Source Title

Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

Volume

8863

First Page

195

Last Page

208

Publication Date

1-1-2014

Abstract

Automatic story generation systems require a collection of commonsense knowledge to generate stories that contain logical and coherent sequences of events appropriate for their intended audience. But manually building and populating a semantic ontology that contains relevant assertions is a tedious task. Crowdsourcing can be used as an approach to quickly amass a large collection of commonsense concepts but requires validation of the quality of the knowledge that has been contributed by the public. Another approach is through relation extraction. This paper discusses the use of GATE and custom extraction rules to automatically extract binary conceptual relations from children’s stories. Evaluation results show that the extractor achieved a very low overall accuracy of only 36% based on precision, recall and F-measure. The use of incomplete and generalized extraction patterns, insufficient text indicators, accuracy of existing tools, and inability to infer and detect implied relations were the major causes of the low accuracy scores. © Springer International Publishing Switzerland 2014.

html

Digitial Object Identifier (DOI)

10.1007/978-3-319-13332-4_16

Disciplines

Computer Sciences | Software Engineering

Keywords

Computer fiction; Computational linguistics; Text data mining; Storytelling

Upload File

wf_no

This document is currently not available here.

Share

COinS