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
Recommended Citation
Samson, B. V., & Ong, E. (2014). Extracting conceptual relations from children’s stories. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), 8863, 195-208. https://doi.org/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