TY - GEN
T1 - Ontology evaluation through text classification
AU - Netzer, Yael
AU - Gabay, David
AU - Adler, Meni
AU - Goldberg, Yoav
AU - Elhadad, Michael
PY - 2009/9/28
Y1 - 2009/9/28
N2 - We present a new method to evaluate a search ontology, which relies on mapping ontology instances to textual documents. On the basis of this mapping, we evaluate the adequacy of ontology relations by measuring their classification potential over the textual documents. This data-driven method provides concrete feedback to ontology maintainers and a quantitative estimation of the functional adequacy of the ontology relations towards search experience improvement. We specifically evaluate whether an ontology relation can help a semantic search engine support exploratory search. We test this ontology evaluation method on an ontology in the Movies domain, that has been acquired semi-automatically from the integration of multiple semi-structured and textual data sources (e.g., IMDb and Wikipedia). We automatically construct a domain corpus from a set of movie instances by crawling the Web for movie reviews (both professional and user reviews). The 1-1 relation between textual documents (reviews) and movie instances in the ontology enables us to translate ontology relations into text classes. We verify that the text classifiers induced by key ontology relations (genre, keywords, actors) achieve high performance and exploit the properties of the learned text classifiers to provide concrete feedback on the ontology. The proposed ontology evaluation method is general and relies on the possibility to automatically align textual documents to ontology instances.
AB - We present a new method to evaluate a search ontology, which relies on mapping ontology instances to textual documents. On the basis of this mapping, we evaluate the adequacy of ontology relations by measuring their classification potential over the textual documents. This data-driven method provides concrete feedback to ontology maintainers and a quantitative estimation of the functional adequacy of the ontology relations towards search experience improvement. We specifically evaluate whether an ontology relation can help a semantic search engine support exploratory search. We test this ontology evaluation method on an ontology in the Movies domain, that has been acquired semi-automatically from the integration of multiple semi-structured and textual data sources (e.g., IMDb and Wikipedia). We automatically construct a domain corpus from a set of movie instances by crawling the Web for movie reviews (both professional and user reviews). The 1-1 relation between textual documents (reviews) and movie instances in the ontology enables us to translate ontology relations into text classes. We verify that the text classifiers induced by key ontology relations (genre, keywords, actors) achieve high performance and exploit the properties of the learned text classifiers to provide concrete feedback on the ontology. The proposed ontology evaluation method is general and relies on the possibility to automatically align textual documents to ontology instances.
UR - http://www.scopus.com/inward/record.url?scp=70349316516&partnerID=8YFLogxK
U2 - 10.1007/978-3-642-03996-6_20
DO - 10.1007/978-3-642-03996-6_20
M3 - Conference contribution
AN - SCOPUS:70349316516
SN - 3642039952
SN - 9783642039959
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 210
EP - 221
BT - Advances in Web and Network Technologies, and Information Management - APWeb/WAIM 2009 International Workshops
T2 - APWeb/WAIM 2009 International Workshops: WCMT 2009, RTBI 2009, DBIR-ENQOIR 2009, PAIS 2009
Y2 - 2 April 2009 through 4 April 2009
ER -