≡ Menu

Live from DAMA – Deploying Semantic Web Solutions: I’ve Built an Ontology, Now What?


Irene and Dean of TopQuadrant (a semantic web company) were up next talking about Deploying Semantic Web Solutions: I’ve Built an Ontology, Now What?

The challenges around Master Data Management have expanded as the number of information sources has exploded. Managing identify, managing semantics across a multi-faceted, distributed world is very complex. The semantic web standards were developed to handle a heterogeneous and complex set of data sources and so are relevant in corporate now that corporations have similar problems.

Focus on Information, Integration and Intelligence (actionable insight) using these standards to access all this different kind of information. Standards like RDF combine with databases, files and more for data storage and must manage versions, evolution, authorship and ownership across these. Four different case studies:

  1. Executive recruitment firm had a large volume of existing data about people including structured information like histories, companies and unstructured like proposals, research briefs, resumes. Internal and external data. They had a very controlled vocabulary and ideas about taxonomies – some 25 different set ups. They wanted flexible search across all these types of data using business terms within their vocabularies. The firm used a hybrid approach that split a semantic query (based on semantic model) into relational database queries and unstructured search engine. They use RDF/OWL to manage the controlled vocabulary and semantic web technology is used to present queries to users intelligently. To build an ontology they used both top-down knowledge acquisition and bottom-up schema analysis – their tool allows the import of a schema so you can treat the schema as part of your ontology.
  2. Major retailer had a web portal for consumers to maintain information about their belongings. Wanted to make it easy for customers to find the products and capture information about these products even though the product lines are very extensive and have different information about those product lines. Need to manage the presentation of this data which also was flexible and extensible. They used an RDF store (Jena SDB) for persistence. Lots of users and have private data for each user plus some of the main catalog (based on MySQL). Logic built on top of RDF uses OWL and rules – these languages are expressed in RDF. Data is metadata in RDF and this meant that the data, metadata and form definitions all use the same language. The ontology came from spreadsheets that described their products – the spreadsheets were scraped for data and then this was combined with a SPARQL query.
  3. NASA have a program to return to the moon and to Mars that involve lots of scientific and engineering disciplines with their own terms, processes, approaches. Communication between these systems was based on XML but each system can have its own view of the truth. Registry, or a federated system of registries, based on RDF and OWL that manages XML feeds to and from the different systems. This gives consistent names and identifiers for global system references.
  4. Pharmaceutical company trying to allow life sciences researchers to explore connections in data about drugs, trials, development etc. For instance a researcher might want to use data about a failed trial to see if the drug could be used for some other purpose. Key thing is to see how different things are connected using semantic web links – RDF graph links. Can use these to navigate around. They needed to pull together lots of Oracle databases into an RDF store on Oracle RDF (like a data warehouse process). This involves transformation and aggregation using a language based on SPARQL to create a warehouse of RDF information. Essential an ETL tool for RDF.

Some of these create a single model of the truth but many do not – the semantic technologies allow modular and linked vocabularies/ontologies. The uniqueness and connectivity are critical aspects of the use of semantic technologies. Ontologies are therefore living structures and evolve and the connections between things are the most interesting element. Ontology-based applications can work with existing systems too.

Interesting discussion and there are clearly ways to use semantic web technologies in decision management. More to come on this topic I am sure.