At the Tampa Bar / Code camp at the University of South Florida, I attended a talk on some really interesting topics. The topics included AI, Machine learning, natural learning, and Data Mining.

The presenter was Don Berndt and he is a professor of MIS at USF.

He had a very interesting story. He told us about the AI winter and how his career in Machine learning was like a boomerang. He introduced the idea of Genetic Algorithms like Bird flocking and how they can be applied to solve technical problems like searching and sorting. He then went on to say his field of study is in combining the power of Machine learned algorithms and Natural Learning.

Artificial life is trying to emulate natural behaviors. If we tightly couple the knowledge natural behaviors with artificial data. We can see incredible optimizations in algorithms used to manipulate that data.

SiteWit the company Don Berndt primarily does his research at is focused on DataMining medical records. They work primarily with the VA hospital here in tampa. Using complex algorithms to organize and search the medical records data is “a big deal now”. Usually these services are accomplished by top down Natural Language Processing algorithms or statistical text mining applications.

Sitewit is attempting to hybrodize these approaches and are seeing some pretty interesting results. NLP is a top down approach and can be very heavy. The worst part is that they usually don’t take in to account abbreviations, short hands, and grammatical errors. Luckily medicine has an amazing Natural learning data source the UMLS. This is the result of a ton of natural learning. The efforts of many qualified doctors and years of hard work have amounted to a network of interconnected medical terms.

Sitewit is injecting the relationships found in UMLS and the power of statistical text search. The main problem is that nothing but frequency counts in statistical analysis.  Adding relationships and weight to the algorithms has increased data mining accuracy by more than ten percent. This following year they hope to see even greater efficiency increases.

This is a field I know nothing about but the talk was extremely interesting. This post is probably riddled with errors and misconceptions. If you have any comments please let me know. I love learning, I just hope that some day I can use this information for something greater than a blog post.

Advertisements