More Related Content
![[WSO2Con EU 2017] WHO CARES? A WSO2 Cloud Oriented Reference Architecture for...](https://cdn.slidesharecdn.com/ss_thumbnails/61krwlpytgis9fu16mrd-signature-10c352b17e3f171dbc0d45b422788a98081eda1dd1fdd2adacfd28644ef1cf59-poli-171107113845-thumbnail.jpg?width=560&fit=bounds)
Similar to Electronic commerce meets the semantic web (20)
Electronic commerce meets the semantic web
- 2. INTRODUCTION: • The intersection of semantic web technologies and business-to-consumer (B2C) e-commerce offers benefits for both online retailers and customers. The authors’ framework highlights why and how the adoption of semantic web technologies can enhance B2C applications and platforms. • As the e-commerce domain matures and expands, many new challenges must be addressed, such as efficient customer engagement, trust management, privacy concerns, and internationalization. • We look at six challenges that pertain to the efficient organization and management of e- commerce (meta)data. Several leading retailers in the e-commerce domain, such as sears and best buy, have already faced these challenges and attempted to address them through community standardization efforts that focus on using semantic web technologies.
- 3. E-COMMERCE CHALLENGES: • The first challenge (C1) is that existing product data are not suitable for automated processing. Rather than have access to structured (meta) data about the products they are offering, many e-commerce vendors receive only free text descriptions. • The second challenge (C2) is that product data often lack interoperability in terms of both syntax and semantics. The lack of syntactic interoperability stems from the fact that different retailers or shops use different schemas to describe products they offer. • The third challenge (C3) is insufficient use of unique product identifiers. Efficient product data integration requires product identity resolution—that is, identifying whether two or more product descriptions, found on different shopping sites, refer to the same product.
- 4. E-COMMERCE CHALLENGES: • The fourth challenge (C4) is the heterogeneity of product category taxonomies. The problem of semantic interoperability is further exacerbated by the diversity of product category schemes or taxonomies used by different shopping sites. • The fifth challenge (C5) is incomplete, inconsistent, or outdated product descriptions. There is often a discrepancy between the richness and variety of product features data offered by manufacturers (known as product master data) and product features available to and exposed by online retailers. As a consequence, product descriptions in online shops tend to be incomplete, inconsistent, or outdated. • The final challenge (C6) is the weakness of current product recommender systems. A great majority of recommender systems in the e-commerce domain are based on collaborative filtering, and thus rely highly on user ratings .
- 6. ADVANCING THE STATE OF THE PRACTICE: • One major advantage of semantic web technologies is an increase in the quality of product data—that is, diversity, completeness, and accuracy. This can be achieved by. • • Gathering and semantically aligning product data from different online retailers, • • exposing product master data on the web as linked (open) data or • • enriching product data with other kinds of data relevant for making purchases—for example, users’ intent and location data. • High-quality product data coupled with semantic technologies for data querying, analysis And reasoning can lead to a number of “tangible” benefits for both online retailers and customers.
- 7. ESSENTIAL SPECIFICATIONS AND STANDARDS: • Table 1 presents standards and technologies that allow for fulfilling the benefits outlined in the previous section. To denote the relationship between a certain technology and an associated benefit, we rely on the following software engineering terminology: • Functional (essential) feature denotes that a feature or technology is essential for the provision of the stated benefit. • Measurable performance improvement indicates that the given feature or technology can quantitatively augment the core benefit. These are cases in which the adoption of the respective technology is not mandatory, but if adopted, it would lead to a measurable increase in the respective benefit. • Quality attribute (...) denotes that a feature or technology, if present, would qualitatively augment the corresponding core benefit, or would augment the system’s quality. This descriptor is used to mark cases in which the core functionality (that is, the benefit) could be provided without the use of the respective technology, but if the technology is used, it could further augment the stated benefit by adding some qualitative difference (such as product search personalization).
- 8. • The fourth challenge (C4) is the heterogeneity of product category taxonomies. The problem of semantic interoperability is further exacerbated by the diversity of product category schemes or taxonomies used by different shopping sites. • The fifth challenge (C5) is incomplete, inconsistent, or outdated product descriptions. There is often a discrepancy between the richness and variety of product features data offered by manufacturers (known as product master data) and product features available to and exposed by online retailers. As a consequence, product descriptions in online shops tend to be incomplete, inconsistent, or outdated. • The final challenge (C6) is the weakness of current product recommender systems. A great majority of recommender systems in the e-commerce domain are based on collaborative filtering, and thus rely highly on user ratings .
- 9. ESSENTIAL SPECIFICATIONS AND STANDARDS: • The bottom-most layer (L1) comprises specifications and standards that are essential for the automated processing of product data published on the web (challenge C1). These include • Standard formats for data exchange on the web, namely XML, javascript object notation (JSON), and resource description framework (RDF); and • Specifications for embedding semantic markup in webpages, including microformats, rdfa, microdata, and JSON for linked data (JSON-LD).
- 10. VOCABULARIES: • The second layer in the technology stack (L2) comprises vocabularies for describing products, offers, and stores. These vocabularies allow for establishing syntactic and semantic interoperability of product data (challenge C2). The most widely used vocabularies include schema.Org and the open graph protocol (OGP). Whereas OGP allows for a very basic product description, Schema.Org offers a detailed vocabulary for the description of various kinds of products, offers, and retailers.
- 11. STRONG PRODUCT IDENTIFIERS: • Strong product identifiers (L3) are required for overcoming the challenge of how to uniquely identify products across different e-commerce websites and the web in general (challenge C3). For some of the outlined benefits (table 1), strong product identifiers are essential features. In other cases, if used, they tend to lead to measurable performance improvements by reducing, if not fully eliminating, incorrect product matching.
- 12. PRODUCT ONTOLOGIES OR CATALOGS: • The use of product ontologies—that is, semantically rich and machine-processable product catalogs (l4)—allows for consistent and unambiguous product descriptions, thus addressing challenges C2 and C4, and leading to high-quality product data. This in turn increases the chance that major search engines will display the given product in the rich snippet format (http://goo.Gl/rxajcm), and also positively influences product visibility in vertical serps. In addition, semantically enriched product descriptions greatly facilitate the provision of advanced customer services such as custom-made offers or product recommendations based on product-specific features, or faceted search based on the features specific to a given product category.
- 13. TECHNOLOGIES FOR DATA STORAGE, SEARCH, AND MANIPULATION: • The fifth layer, semantic technologies for storage, search, and manipulation of product data (L5) includes rdf-based technologies (RDF schema and the web ontology language) for explicit representation of data semantics, which in turn allows for sophisticated search and automated reasoning over the available data. Search is powered by the SPARQL data query and manipulation language, which allows for advanced, semantic search of stored data. Today’s RDF triple stores, some of which are shown in figure 1 (L5), support enterprise-level data integration and sophisticated cross-department or sector data search and reasoning. • One of the main advantages of data stores over traditional relational databases is in the flexibility of the underlying data model.
- 14. INTELLIGENT E-COMMERCE APPLICATIONS: • At the top of the technology stack (L6) are intelligent e-commerce applications that use technologies from the lower layers (L1–L5) to offer retailers or customers the previously described benefits. These applications might feature, for instance, faceted search at web scale,3 search of long-tail or highly customizable products or personalized product recommendations. applications of this type have only recently started to emerge, which is expected considering the novelty of the underlying technologies.