Author:
Steven B. Kraines
Affiliation:
The University of Tokyo, Japan
Keyword(s):
Knowledge Representation, SKOS Ontologies, Semantic Similarity.
Related
Ontology
Subjects/Areas/Topics:
Applications and Case-studies
;
Artificial Intelligence
;
Data Engineering
;
Enterprise Information Systems
;
Information Systems Analysis and Specification
;
Knowledge Engineering and Ontology Development
;
Knowledge Reengineering
;
Knowledge Representation
;
Knowledge-Based Systems
;
Ontologies and the Semantic Web
;
Ontology Engineering
;
Symbolic Systems
Abstract:
The effect of additional domain knowledge provided by a SKOS ontology on the accuracy of semantic similarity calculated from product item lists in purchase orders for a manufacturer of modular building parts is examined. The accuracy of the calculated semantic similarities is evaluated against attribute information of the purchase orders, under the assumption that orders with similar attributes, such as the industrial type of the purchasing entities and the type of application of the modular building, will have similar lists of items. When all attributes of the purchase orders are weighted equally, the SKOS ontology does not appear to increase the accuracy of the calculated item list similarities. However, when only the two attributes that give the highest correlation to item list similarity values are used, the strongest correlation between item list similarity and entity attribute similarity is obtained when the SKOS-ontology is included in the calculation. Still, even the best
correlation between item list and entity attribute similarities yields a correlation coefficient of less than 0.01. It is suggested that inclusion of semantic knowledge about the relationship between the set of items in the purchase orders, e.g. via the use of description logics, might increase the accuracy of the calculated semantic similarity values.
(More)