Guiding Eyes with Watson predicts dog raising success
Cognitive computing helps the guide dog organization match dogs to trainers
Training guide dogs can be a tricky business. Even for one of the largest and most successful guide dog training organizations in the world, Guiding Eyes for the Blind in Yorktown Heights, NYC, only just over 1/3 of the dogs successfully make it through the training program and can be successful paired with a blind person.
So the role of Jane Russenberger, Director of Genetics and Breeding for Guiding Eyes is key. She keeps information about the genetics of every dog. She has info about their mother and father, brothers and sisters. She knows about the first two years of their life, from the initial puppy testing and through the training program. And crucially, she knows whether they are successful or not.
So earlier this year, we linked Jane with Professor Chris Tseng at San Jose State University (SJSU) on a project to address the question: can the data hold clues to the reasons why some dogs make guide dogs and other don’t.
The professor employed IBM Watson’s cognitive capabilities to analyze the masses of structured and unstructured data from Jane and her team. After only 3 months, Professor Tseng had managed to build a model that was 100% accurate at predicting whether a dog would successfully complete the program and was 100% accurate at running similar predictions among their pool of over 500 volunteer trainers. Not bad for first time out of the gate.
Is the work done? Nope, we’re only beginning.
Starting this Fall Jane is looking to take 500k pieces of data on the dogs measuring traits as varied as their reaction to noise, excitability, how confident a dog is with stairs, the list goes on. The plan is to take all that data and feed it to Watson to identify the combination of traits that help a dog be successful.
But Jane’s vision extends beyond this. She would like to share data across the 80 guide dog schools worldwide to uplift the performance of all programs, even those that currently don’t have breeding programs. She believes there are hidden keys at the molecular genetic level that only a system like Watson can uncover.
See CNBC coverage of this story and learn more about the origins of this project. If you are at World of Watson this year, you can talk to Jane (oh, and a guide dog) for yourself in the cloud area of the expo hall.
