Flexible Classification Standards for Product Data Exchange

One essential part of e-business is the exchange of product data between business partners. Classifications have been developed as a means to clearly describe the semantics of product descriptions. They provide schema elements like properties and classes and define their semantics by some formal relationships and some textual (informal) definitions. This chapter gives an overview about the modelling levels which have to be considered for the development and use of classifications. In addition, it introduces briefly ISO 13584 (PLIB) as one important data model for classifications and it characterizes a number of classifications which are used in today‟s product exchange processes. Many of the classification standards use a quite primitive data model which leads to problems in their use and maintenance. By exploiting some features of more advanced data models, many of these problems can be overcome. We propose the introduction of additional class hierarchies as an example for adding more flexibility to a classification and discuss this proposal in the context of eCl@ss, an important European classification. INTRODUCTION A fundamental requirement for implementing full e-business is the ability to exchange product information between different business partners and their software systems. As far as productnumbers, prices, delivery information, etc. are concerned, exchange is possible on the basis of general models which are applicable for any kind of product. But the technical diversity of products leads to a diversity of technical descriptions. This makes it impossible to define a general model covering all aspects of all types of products in a concise way: The description of e.g. bolts and nuts requires fundamentally different information than the description of integrated circuits or refrigerators. Whereas STEP, a series of standards on product modeling (International Standardization Organisation [ISO], 1994), defines a big number of models for various domains to describe e.g. geometry or other representation models of a product, the exchange of technical product information in e-business and e-engineering is basically done by describing products by their characteristics or properties. This information exchange is normally based on a meta data approach: Information about a property of a product is exchanged as a pair (property_ref, property_value), where the property_ref is an identifier of a concept in a classification (often also called a dictionary or product ontology). For the correct interpretation of the property_value, the receiving system has to refer to the classification, where the meaning of the property is defined (textually and possibly supported by graphical means) and further information is available like names (in different languages), synonyms, data type, unit of measure, relationship to other concepts, etc. Thus, we have a very simple structure for the exchange, but we need classifications as additional resources which are referenced from the exchange structure. The consistency and accuracy of the elements of these classifications is important for users: If the resources of a classification standard are not able to describe their products appropriately they will not be able to transmit the required information and to use the classification standards in their product data exchange processes. Most classification standards are developed and maintained by a consensual process which is based on industrial working groups which consider the change requests from users and decide on their integration into the existing classification standard. The difficulties of these activities should not be underestimated, in particular in view of the growth of the classification systems over time. Today, classification standards often consist of more than 10,000 classes, several thousands of properties, and an even bigger number of class-property relations. Due to the increase of product areas which are being integrated in product classifications and the increase of requirements of applications, the growth of classification standards is enormous. For example, Table 1 shows the development of the number of classes and properties over the last versions of eCl@ss (figures obtained in personal communications with members of the eCl@ss-office, some information can also be obtained from CEN, 2010). V4.0 V4.1 V5.0 V5.1 V5.1.1 V6.0 V7.0 Publication Date 2001/08 2002/09 2003/12 2004/09 2005/09 2008/04 2011/02 Commodity Classes 10,190 12,565 20,379 21,100 22,203 32,590 37,868 Properties 2,303 5,504 3,667 5,525 6,941 10,930 15,397 Table 1: Growth of Classes and Properties in eCl@ss Thus, current classification standards have to face a number of problems: The classification standards are growing, so that the organization of the maintenance process becomes more difficult. Different user groups put different requirements on the classification standards. Classification standards are based on simple data models which put limitations on the expressiveness and flexibility of classification standards. Based on an overview of classification models and standards, we will argue in this chapter that future classification standards have to provide different views on a common set of basic classes and properties. This will allow for a better support of conflicting requirements, and it will also support the maintainability of big classification standards. Basis for such an improvement of classification standards is the use of data models which provide sufficient resources to build a more flexible classification standard. Our example is eCl@ss, a horizontal classification standard with the approach to cover all industrial products in all phases of their lifecycle (see www.eclass.eu). We will illustrate for eCl@ss that by means of the capabilities of its new data model multiple classification hierarchies can co-exist. This will result in a number of benefits for its maintenance and use, both by reducing the number of class-property relationships and by increasing the consistency of the classification standard. The rest of this chapter is organized as follows: In section 2 we give an overview about classification standards and their capabilities to describe and organize product information. We introduce the standard ISO 13584-42 (ISO, 2010), also called PLIB, as a data model which is increasingly used for classification standards and show how eCl@ss has adopted this data model in its recent version 7.0. In section 3, we highlight some of the problems which result from deficiencies of the simplistic data models. Section 4 describes the definition of additional class hierarchies for a classification to overcome these problems, and it contains some calculations for our example eCl@ss about the reductions of property-class assignments. Finally, in section 5 we give a summary and an outlook on further work. CLASSIFICATION STANDARDS AND THEIR CAPABILITIES Fundamentals of classification standards For describing product data which can be exchanged across different partners and their systems, the meaning and organization of data in exchange files or messages needs to be specified. This is done by defining a model or a schema to which the objects in the exchange file can be related to describe their semantics. Such models are called classifications in this chapter. Unfortunately, this terminology is not very clear and in different contexts a variety of names can be found having a similar meaning. Examples of such models, which also give an impression about the variety of terms, are E-business classification standards like eCl@ss (www.eclass.eu), ETIM (www.etim.org), Proficl@ss (www.proficlass.org), and UNSPSC (www.unspsc.org), ISO and IEC standards providing product dictionaries like ISO 13584-511 (ISO, 2006) about fasteners, ISO 23584 (ISO 2009) about optical devices, and IEC 61987 (IEC, 2009b) about control instruments, Reference libraries like ISO 15926-4 (ISO 2007), RosettaNet Technical Dictionary (RNTD) for electronic components (RosettaNet, 2007). Classifications specify sets of concepts which are required to describe the products of a domain or of several domains. For instance, ISO 13584-511 (ISO, 2006) contains various classes of fasteners together with their properties which can be used to describe a specific fastener in a catalogue. As a matter of fact, these classifications provide different kinds of information about the products of their domain. Some examples: ISO 13584-511 (ISO, 2006), ISO 23584 (ISO, 2009) and IEC 61360 (IEC, 2009a) organize their product classes in a hierarchy which is driven by property inheritance, i.e. for upper classes in the hierarchy, properties are defined which are inherited by subclasses. Other classifications do not provide a class hierarchy, e.g. NE 100 (International User Association for Automation in Process Industries [NAMUR]), 2010), OTD (ISO, 2010b), and RNTD (RosettaNet, 2007). Sometimes, a hierarchy of classes is provided which is not based on property inheritance but on market segments (eCl@ss, UNSPSC, the Global Product Classification of GS1 (GPC, see http://www.gs1.org/gdsn/gpc), and the Common Procurement Vocabulary (CPV, European Commission, 2008). ISO 15926-4 (ISO 2007b) only defines the classes and properties but does not relate them to each other. Thus, the classifications provide different facets or features as a means to actually describe products in an exchange (or in a database). Which features they provide is based on two (not independent) elements: 1. The intention of the provider: For what purposes is the classification supposed to be used? 2. The capabilities of the underlying data model which provides the means for building classification. Based on their intentions, the classification providers will decide whether they need a class hierarchy, how the class hierarchy is organized, how classes and properties are related, etc. But if the providers have selected a data model they are bound t