Towards a Semiotic Metrics Suite for Product Ontology Evaluation

Introduction

Product-related information is of paramount importance in many interorganizational applications, since it concerns goods and services being procured, manufactured and sold to customers. Due to the involvement of multiple organizations, there is a need for integrating product-related information, e.g., by standardization or mediation. In the past years, product ontology has attracted both industry and academia because of its potential contribution to solving integration problems (Shim & Shim 2006). A product ontology provides, at least to some extent, consensual definitions of concepts and inter-relationships between these concepts in a product domain of interest. Most product ontologies define a hierarchy of product classes and respective properties for describing product instances. Such ontologies may support finding and comparing products being offered by multiple suppliers and described in distributed data sources, or allow for benchmarking the procurement activities of organizational units (Doring et al., 2006). Ontology users are required to annotate their product instance data accordingly.

Product ontologies have already emerged in diverse industries and for various tasks (Park et al., 2008). However, assessing the quality and suitability of a given product ontology, i.e., to what degree it actually meets user requirements, remains a critical question for potential ontology adopters. This question is the focus of ontology evaluation, which aims at providing metrics reflecting the ontology’s quality and suitability. There is great difficulty in determining what elements of quality to evaluate. In other words, what factors should be considered in evaluating product ontology quality? Current research yields a number of approaches, metrics, and tools for automatically evaluating ontologies (Garcia-Castro et al., 2007; Hartmann, 2005). However, most of this research originates from the Semantic Web arena, and therefore relies mainly on the expressiveness of ontology languages such as DAML (DARPA Agent Markup Language) and OWL (Ontology Web Language); hence their scope is constrained by these languages and does not take the specific setting of product ontology into account.

Very often, an ontology is regarded as an artifact used by a community as a common vocabulary without considering the organizational properties of the respective community and thus the inter-relations within the community (Zhdanova et al., 2007). For example, a community that often uses product ontologies is made of entities belonging to a supply chain. A supply chain is a system of entities participating in producing, transforming, and distributing goods and services from supply to demand. A single product ontology is thus used within supply chains and determining its quality and suitability has to consider the supply chain characteristics, e.g., by distinguishing different roles such as manufacturer and distributor. A major trend affecting supply chains is individualization, caused by customers demanding individualized products, which are tailored to their specific needs (e.g., custom-made products) (Coates, 1995) (Kirn, 2008). For instance, enabling customers to order custom-made shoes via an e-commerce application does not only concern the e-commerce firm but also the stakeholders in the respective supply chain (e.g., manufacturer and its suppliers). Here, a product ontology may help provide a common terminology and means of describing products along the entire supply chain.