June 2, 2007 8:27 AM
Semantic Technology skepticism and optimism
Until now we have discussed the architecture of a semantic web application open market. This is being summarized here with a different example:
| Natural Language Sentence | Query |
| Natural Language Grammar Interpreter | Term recognition (noun, pronoun, verb, adverb, adjective) |
| SPARQL | Logic formation |
| OWL, OWL-S, RDF, RDF-S | Machine-interpretable ontology & meaning of a thing <subject, predicate, object> |
| XML | Machine-interpretable description of 'thing' |
| IRI | Web resource address on World Wide Web |
| thing | Web resource |
The intelligence of an inference engine that works on a natural language query to transform it into machine-language representation (RDF/OWL and SPARQL) is in identifying the natural language grammar terms in the sentence. The noun is taken as subject/object and verb as property. What to do with pronouns, adverbs and adjectives? There may be more than one interpretation of a term, adverb cheap can be mapped to ascending order search for cost, price, charge or money. The pronoun she/her must be mapped to search for a female 'thing'.
| Example: “good furniture” good is an adjective and furniture is noun. What property of furniture search engine must look for? It may be low cost and high quality. While low cost is an ascending search on cost, quality can be determined by many factors such as: wood and workmanship. A SPARQL query furniture with low cost and wood type is formed. This is converted to RDF triples <furniture, cost, value> and <furniture, wood, type>. Here the search engine will inform the web user about the URI of the suppliers of good furniture. The inference engine may convert the natural language query into other SPARQL queries such as: high cost, low cost, high quality, or specific wood furniture. |
Tools like Protégé may be used to convert the query into machine interpretable ontology expression (OWL/RDF). Oracle Spatial 10g is another data management tool for RDF based applications. The semantic query is submitted to the semantic web service application to get the information. We have discussed one type of semantic application: context mediator. This semantic application has also been called semantic broker in this white paper: "The CIO's Guide to Semantics Version 2". In this white paper Dave McComb has also discussed the semantic web standards:
- Standards to express semantics – these are the normative specifications (such as: RDF, OWL, SPARQL and XML) that are not specific to any application or industry and are used in all semantic web applications.
- Domain specific semantic standards – Application, domain or industry specific standards to standardize the meaning or terms and specific vocabulary, such as Dublin Core, RSS, FoaF and UMLS.
Dave McComb has also discussed the effort required for resolving semantic and syntactic differences in data from different sources. He has described semantic broker architecture for resolution of these issues. Semantic broker is an application to resolve semantically tagged messages for interoperation between semantic applications. This architecture is same as what we discussed earlier. Stephen Downes is not very optimistic about semantic web, he has raised three issues (standards, vocabulary and APIs) that we have discussed on IT Gumbo.
I am going to be blogging live from a couple of days of the 
Leave a comment