Getting Started with Semantic Technologies

[Note: I originally posted this on my personal blog in May 2008, but it's so much more appropriate here!]

I met a gentleman who was at SemTech for the second time. He was very excited, really jazzed about the whole experience, but still wasn't sure how to win people over back at the office. How could he get them to embrace semantic technologies? It was a hard sell.

I wasn't surprised at all; it's the same song Kevin and I sang when we worked together. We got some great advice from people like Andy Shain and Eric Miller, which I have heard repeated at SemTech each year since.

Take baby steps. Keep it simple.

Don't try to boil the ocean. Don't try to sell a complicated new method or tool to your team. Simply consider the problem at hand, and, if there is a semantic technology that can solve the problem, try it as one of the development options.

Need access to employee contact information, and want it in RDF? Don't ask everyone to generate a FOAF file, or try to convince your directory team to convert to a triple store. Ask if you're allowed to get a feed from the LDAP system, and if so, ask for a copy of the schema. Convert the feed rather than the source.

Want your team's Sharepoint document repository to use Dublin Core? Set the columns up in the appropriate format, so that when you export from Sharepoint to Excel you'll have it in a format ready to be converted to RDF by one of many available tools. (See Simile)

These may sound like hacks, but SHOWING the outcome of it will help to win people over to more sophisticated implementations rather than trying to TALK them into supporting your efforts. Start small and only attempt to learn advanced topics when you find a need for them. Trying to learn OWL by sitting down and reading it could be a frustrating exercise. Start with simple models, learn about domains, ranges, and how to 'disjoint' only when you find that your model needs it.

One useful thing that I learned while working at Intuit was to separate for people the What from the How. Talk to skeptics about the What, talk to co-conspirators about the How. Spend what free time you may have prepping elevator pitches for the What, and learning about the How.

Most importantly, have fun. It's fascinating to see how each new source we add connects to the others, building the graph in unexpected ways.