Bob Renner: In one example and use case, we married our sales data with Twitter feeds so that we can access the API’s. We pull the data in, normalize and correlate it, and we created a dashboard that allowed our clients to look at sentiment. We were able to dimension that along with the sales data. We product a very different, simultaneous view of geography, volume of sales, and demographics of who they’re selling to and then generalize sentiment about how people are talking about the data from social media standpoint. The interesting part of that is it uses some forms of natural language processing and parsing to take free-form data and turn it into structured data from the Twitter feeds and marry it up to other structured data.
Sramana Mitra: Who provides that unstructured data to structure data transformation technology?
Bob Renner: That was our technology using generally available applications for parsing that data.
Sramana Mitra: In each of these cases, obviously you’re dealing with disparate types of systems that you are integrating. Could you simplify the different types of systems that you’re working across or bringing together?
Bob Renner: We have never met a system that we didn’t like or we couldn’t integrate to. We, literally, have tens of thousands of different clients with a variety of hundred thousand different permutations and combinations of systems that we interface with. The strength of the Liaison platform is that our semantic integration technology has four issued US patents.
We are built on the premise that we believe standardization at the end point is not only unlikely, but it’s impractical. A lot of what we do and the strength of our technology is doing that data transformation, translation, and adaption to these systems. That’s one of the first things that we relied on in our development. It was the premise that we needed a ubiquitous tool that could translate and transform, and adapt to any system or data.
Sramana Mitra: If you look the data integration world today, we are working with a lot of different types of systems and data. Unstructured data presence in business today is much higher than it used to be because of social media. Give me a couple of really hairy integration problems that you see out there that have not been solved yet.
Bob Renner: You mentioned unstructured data. I think everybody gravitates towards that. To be honest with you, a lot of the structured data problems are yet to be solved on a large scale. What I mean by that is I think the combination of accessing authoritative data – profiling the data to determine whether there are redundancies, which data source for common or overlapping data is, in fact, the best authoritative data source. I think sorting so that you’re giving the best possible and the most up-to-date view of the accessible structured data is a pretty hairy problem generically. We’ve actually been faced with that at many clients. Trying to pull out one specific example might be difficult. But I’d say generally, inside an enterprise or ecosystem, what’s surprising to me is the amount of redundant data and the challenge to profile that data and determine which source should be used in your final product.