A '72 Chevy Nova, Unstructured Data and Cognitive Computing
March 16, 2017
Earlier in the week, Genus Technologies sponsored Insurance Nexus in Chicago. If one trend emerged from the conversations we had there, it's that for insurance companies, unstructured data and cognitive computing represent the next frontier to be tamed, if not conquered. Take for example all of the information contained in a simple photograph of a classic car. That information holds answers to many questions of particular importance to an insurance company quoting a policy including:
- The year, make and model
- Whether or not the paint is an original color
- Whether or not the vehicle is original or modified
- Whether or not the vehicle has been repaired
- The overall condition of the exterior
Unfortunately, all of data that provides this information is unstructured. It lacks a pre-defined data model. To assess this particular vehicle in order to write a policy, the insurer would need the skills and talent of an experienced classic car appraiser who could look at the photo and surmise:
- Shape of the grill, headlights and front bumper assembly confirm the vehicle is a 1972 Chevrolet Nova
- Cranberry red was a factory-offered paint color in 1972 and looks to be correct
- SS badging appears to be accurate and original supporting the claim that this is a rare Nova "SS"
- The alignment of the body panels suggest the car has not been damaged or repaired
- Overall condition of the exterior shows signs of obvious care and support the owners claim that it is original
That approach is so 2016.
Cognitive computing is here, and it’s transforming the value and usage of unstructured data. Self-learning systems, such as IBM Watson, that use data mining, pattern recognition, and even natural language processing are augmenting the need for human engagement as in the above example. With cognitive computing, machines can mimic the way our classic car appraiser’s brain works. And this is only one simple example.
So as you look at the way your organization functions, where do you see a fit for cognitive computing? How can unstructured data become actionable through the use cognitive systems within your organization?
Submitted by Glenn Seaberg