“People want to be a part of something that is the future”
Marc Benioff, Behind the Cloud
Anyone who has just discovered the concept of Semantic Web wonders why, despite the apparent simplicity and usefulness of the concept, its deployment takes such a long time and still does not produce any noticeable effects on business. Why a framework that allows data to be shared and reused across application, enterprise, and community boundaries remains only a concept rather than a dominating standard?
The following reasons are common to explain the slow acceptance:
- Complexity and difficultness of semantic markup;
- Risk that semantic markup could facilitate automatic data retrieval and decrease site traffic;
- Practical impossibility of a universal ontology.
Today, one of the best practical implementation of Semantic Web is Shema.org supported by major search engines like Google, Bing, Yahoo, and Yandex. Shema.org simplicity determines its success, but limits its practical application, especially for product descriptions. Product template is missing the concept of the “category”, for all products used single common template. Moreover, multilinguality is not explicitly supported by Schema.org microdata. Microdata is rather a tool for nice snippets in search results, rather than a solution for machine information processing.
To share product information in Semantic Web we need tens of thousands of product templates. The product ontology should be publicly available and generally accepted. Until today, product templates were available mostly on commercial basis. The actual cost of high-quality product templates (identification guides) is high enough, it can be several thousand dollars per category. It is necessary to involve professionals, industry experts who are knowledgeable in a particular category of products.
Today such a scheme does not exist for complex technical products. GS1 and IceCat could be used as a basic for consumer products, and eCl@ss for B2B products and services. eCl@ss offers a set of attributes for the categories, but their number severely limited. Strict hierarchy and lack of inheritance of attributes limits its application to develop and manage product templates.
We face a number of challenges when creating product classifier, just to mention a few:
- National standards. There are different national naming standards for products. Attributes values might be standardized and national standards provide different possible values.
- Product variants. Some products have multiple variants; moreover, some characteristics could be defined only by order. We need a policy to define what a product is and what a product variant is.
- Synchronization of changes in the products when modifying the product template;
- Management of attributes to ensure sufficient flexibility and completeness;
Identification of categories and feature selection to create subcategories.
Sometimes we found an answer in international standards like ISO 22745, ISO 8000 and 29002.
We believe that presenting product ontology and product data in RDF schema and OWL will allow to practically implement the Semantic Web concept. By this time may be it will not look likes Tim Berners-Lee originally envisioned. Any visual formatting is not required for automated exchange, so we might have two types of web-sites – for people and for machines, or, alternativly, for automated data echange API with semantic markup support could be used/