How are semantic modeling techniques changing data modeling, and what role do you think data modeling will play within companies in the near and distant future? Will there still be a need to model data?
Data modeling will be with us for a long time to come – but the form that data modeling takes is changing and will continue to evolve. Using semantic technologies, we’re able to understand and leverage our data in new ways to automate knowledge discovery. What does this have to do with data modeling? EVERYTHING!
Semantic technologies (e.g., the semantic Web, text mining) rely on ontologies, which I call “data models on steroids.” Many of the same principles of data modeling apply (more or less) to semantic modeling for ontology development. For example, as data modelers we are (or should be) extremely concerned and interested in relationships between data entities. What is a data entity (especially in a conceptual data model)? They typically are the nouns of our business: customer, product, location. In semantic modeling, we are interested in how terms relate in taxonomies and thesauri. This equates well with parent-child, subtype and associative relationships in data modeling.
Of course, not every platform will be able to use semantic models in the near term (how many systems are still out there on mainframes running VSAM?), and semantic models are not going to be able to address every problem.
Some propose that with advances in hardware and software, it will no longer be necessary to model data for optimal performance – for example, in a star schema. There is a grain of truth to this, but I think some pretty big assumptions are being made. Sure, you could throw all of your data onto a data warehouse appliance as is and let the business intelligence (BI) semantic layer hide all the complexities of navigating the normalized data. Without the pesky data modeling effort that’s usually required, you likely could bring up the data warehouse more rapidly. And of course, appliance vendors would love to sell you as many server blades as possible. However, a star schema model is still quite simply the fastest structure for querying large volumes of data.
Regardless of the technological advances that occur, there will at a very minimum continue to be a need for people who are able to understand business needs for data and can model and define that data accordingly, even if it’s done solely at the conceptual or semantic level. Data and semantic modeling require specific skill sets, and not everyone is able to deal with the level of abstraction required for modeling. Seeking a career in data and semantic modeling is not an unwise move!
This was first published in June 2011