Relation-Centric Semantic Annotation using Semantic Role Labeling and Coreference Resolution
BibTeX
@MISC{Lin_relation-centricsemantic,
author = {Chia-hung Lin and Sheng-hao Hung and Chien-hsiang Liao},
title = {Relation-Centric Semantic Annotation using Semantic Role Labeling and Coreference Resolution},
year = {}
}
Bookmark
 |
 |
 |
 |
 |
 |
OpenURL
Abstract
Abstract- Automatic semantic annotation based on domain-specific ontologies is a one of the critical issues for the success of the semantic web. Most existing approaches focused on the detection of concepts such as named entities, dates, monetary amounts. This study explores automatic semantic annotation techniques for applications using relation-centric ontologies which represent domain knowledge using a set of concepts with many inter-class relations. We propose a framework to detect event-based concepts and inter-concept relations using semantic role labeling and coreference resolution techniques. We gave an illustration of the processes by a semantic annotation application using CIDOC-CRM as the underlying ontology. Experiments using archives with a large number of image descriptions were conducted. The primitive results show that the accuracy is about 80% or so.
