Semantic Technology Limitations and Criticisms

While the above discussion provides support for the use of Semantic Web technologies – and it is our view that it is currently the most promising solution available – it is no panacea. In particular there are a number of important features where it is limited, such as lack of native support for representing and reasoning over dynamic knowledge, lack of enterprise-grade tooling, and slow adoption in the developer community.

There is a significant gap between the currently available Semantic Web technologies and the need for native temporal data modelling and reasoning. In addition to “home- brew” strategies that use the standard RDF data model (SYD10), extensions to the model semantics and the SPARQL query languages have been proposed (LLPG05; GHV07; PUS08; TB09; EWK90), however, there is yet to be movement towards the development of standards.

Another issue is that state-of-the-art Semantic Web technologies are designed to handle large volumes of static data, while pervasive systems bring the need to work with highly dynamic data, often with real-time reasoning and query responsiveness require- ments (SYD10). A consequence is that the majority of off-the-shelf reasoners do not provide support for deletion of facts after their insertion, requiring all inferences to be re-computed after any operations on the knowledge base. The Pellet reasoner (SPG+07) is one notable exception that provides support for handling knowledge base updates incrementally.

Technologically, the readability and accessibility of serialised RDF data models has been criticised. The ‘standard’ serialisation, RDF/XML, is verbose, not human friendly, requires RDF’s terminology (of resources, properties, literals and blank nodes) to be recast as XML elements and attributes, and requires a graph to be encoded as a tree. However, this is primarily an issue of implementation, rather than a criticism of the model itself. The development of alternative ‘pretty’ syntaxes (for example N3 (BLC11)), although not yet endorsed as W3C standards, alleviate this concern and are widely understood by tooling.

to developers, and the commitment required by developers and companies to adopt them. Considerable familiarity with namespace semantics, ontologies, reasoners, and the vocabulary and associated semantics of RDF, RDFS, and OWL are required before development of simple systems can begin. The marketplace for assistive ontology development tools is currently small, with only a few choices available for developing and validating RDF/OWL ontologies. Protégé (NFM00), TopBraid Composer (top), and NeOn Toolkit (neo) are among the most prominent.

The decision to adopt Semantic Web technologies, whether taken by an individual or organisation, comes as a tradeoff between its advantages and risks. The ability to define a structure for knowledge that exactly matches a chosen sub-domain, to describe the richness of this structure, to have it compose cleanly with other such descriptions of complementary sub-domains defined independently – and to be able to exchange all this knowledge with anyone on the web is clearly desirable. However, the current reality is far from this vision. Adopters are buying into a vision with the hope or expectation others will follow suit, without it being clear to what extent that this will be the case. In December 2013, the W3C announced that all Semantic Web Activities are being superseded by the Data Activity, with a “focus on deployment and integration within the broader landscape” (Zai13).

Should it be the case that the Semantic Web is not realised using the currently proposed technology suite, it is likely that any successor technologies will support many of the desirable features of RDF and OWL, and will provide migration paths. For example, recent emphasis on JSON in the web community has given rise to JSON-LD, a light- weight Linked Data format based on JSON that aims to help JSON data interoperate at Web-scale (Lan13). The JSON-LD specification contains the concept of a “context”, which provides basic interoperability with the RDF model by allowing object properties in a JSON document to be linked to properties defined in an ontology.

3.1.10

Summary

The suite of Semantic Web technologies has the potential to play an important role in knowledge management of pervasive systems from the perspectives of modelling and representation, inspection and manipulation, discovery and distribution, and support for integrating with legacy data sources and technology. The technology suite is expressive and flexible, although adoption requires mastering a steep learning curve.

Most is to be gained from the Semantic Web at the point where it is fully realised – adopted by the masses – and there is a distinct possibility that this may not come to fruition. However, even if only partially realised, the Semantic Web provides the ability

to capture a set of data and metadata, to reason on the captured information, and to openly exchange said data and model in a domain neutral way. These features improve upon the interoperability and semantic clarity over alternative technologies, supporting the view that developing pervasive systems using Semantic Web technologies remains a promising approach.