Chapter 2: Frameworks and Reference Models

Title: Semantic combination of matching methods for product data interoperability

Author(s): Il Yeo, Lalit Patil, Debasish Dutta

Address: Department of Mechanical Engineering University of Michigan 2350 Hayward St., 2250 G.G. Brown Ann Arbor, MI 48109, USA | Department of Mechanical Engineering University of Michigan 2350 Hayward St., 2250 G.G. Brown Ann Arbor, MI 48109, USA | Department of Mechanical Engineering University of Michigan 2350 Hayward St., 2250 G.G. Brown Ann Arbor, MI 48109, USA

Reference: International Conference on Product Lifecycle Management 2008 pp. 167 - 176

Abstract/Summary: Determining semantic correspondences across heterogeneous product representations is critical for seamless translation and, thereby, for integration in PLM. Multiple attributes are used to determine semantic similarities using different metrics or methods. However, there is no formal basis to define a correct approach, and the results are not reliable. Therefore, there is a need to consider a variety of features along with the variety of methods to approximate similarity. In this paper, we demonstrate the need for nonlinear combination of the matching approaches to reduce uncertainties in semantic mapping. We demonstrate the use of Support Vector Regression (SVR), to capture the nonlinear interrelationships and find semantic maps between product assembly ontologies in disparate representations.

Order a copy of this article