Tennessee Tech University Department of Computer Science Blog

(This post was authored by Dr. Doug Talbert.)

I have a student who occasionally wears a “Schoolhouse Rock” shirt evoking some nostalgia for the short, educational cartoons that were shown on television when I was a child. The catchphrase for Schoolhouse Rock was “Knowledge is Power,” and even though it has been a long time since those were regularly aired, the truth of that catchphrase still resonates with me today, and as a data scientist, my job is to unlock the power of knowledge that is hidden inside data. In fact, Yann LeCun, Director of AI Research at Facebook, once said, “Most of the knowledge in the world in the future is going to be extracted by machines and will reside in machines.”

A group of TTU computer science students and I are trying to enable machines to extract knowledge from medical data and turn that knowledge into power in the fight against cancer. We have partnered with researchers at Oak Ridge National Laboratory and the Medical University of South Carolina to seek knowledge that will improve our understanding of cancer-related disease processes, enable more informed decisions regarding prevention and treatment, and possibly assist in the development of new prevention and treatment options.

This complex problem presents many challenges to both data scientists and clinicians. The relevant data includes electronic health records, clinical reports written in natural language, and genetic data containing sequences of thousands of genes. Our work will involve using data science to transform that data into something the computer can manipulate, to link related data together, to identify meaningful patterns, and, ultimately, to translate those patterns into knowledge that can help us prevent and treat cancer more effectively.

So far, I have focused mostly on the medical goals of our project – a better understanding of cancer and improved cancer prevention and treatment. We also hope to advance the state of knowledge in data science/machine learning. Our machine learning goals have been inspired by recent advances that seek to equip computers with more human-like learning capabilities – an ability to learn continuously, an ability to autonomously apply knowledge learned in one context to improve learning in another context, and an ability to self-direct learning.

The development of such an advanced learning system in an area as complex as cancer research will only become a reality through many small steps over a long period of time. Like the system we plan to build, we’ll learn as we go, and, hopefully, along the way, we’ll turn data into power that helps us beat cancer.

(This post was authored by Dr. Bill Eberle)

When I tell people I worked on Star Wars, I generally get that look of “What?”, “Really?”, or “You’re too short for a storm trooper.” I then have to explain to them that it was actually called the Star Wars Defense Initiative – a national defense system created during the Reagan years to protect our country from a potential Russian nuclear strike.  It was fun working on the project.  I got to work in what is called a Tempest building – a building so thick that signals cannot get in or out.  I was using high-end graphical work stations, and programming fairly complex mathematical equations.  I even got to meet senior military personnel, whom I would then demo simulations of the defense system.  Unfortunately, I never got to meet Admiral Ackbar.

While all of this work was satisfying, I was not getting to use any of what I had learned over the last several years while I earned my Masters’ degree in Computer Science with a concentration in Artificial Intelligence. Fortunately, an opportunity presented itself. A local division of a major telecommunications company was starting a new project, and was looking to hire software engineers on a project involving marketing data.  While, at that time, I didn’t see the A.I. in what was being advertised, something intrigued me about the opportunity.  So I applied.

I got a call a couple of weeks later to meet with the hiring manager. While we talked about the project, I realized that the general problem they were trying to solve was handling lots of data (nobody called it “big data” back then) and creating tools that would provide knowledge to their users – or, today, we would call it data science.  I then proposed to him some ideas out of artificial intelligence that could be used to analyze the data and present information in a way that would be easy to use and understand by their customers.  I think that sold him, because he called me back the next week and asked when I could start.

I then got to spend the next three years doing data science! I was involved in everything from natural language processing, to data warehousing, to complex SQL querying of the data, to visualization of the data.  My MS thesis had been in natural language processing, so I used that expertise to create software that allowed marketing people to ask the system queries, in English – like how many people are married, drive a BMW, have a dog, and 2.5 kids – and in turn it would generate SQL queries to the data warehouse.  Results were then translated back into English (or into a table, if they preferred), making it very easy for them to understand the results.  And all of this was done on a very big data warehouse, which, at that time, was actually the largest data warehouse in the world.

In the end, it was all about creating software that made it easier for users to understand their data.

I may not have been able to destroy the Death Star, but it was still lots of fun.

Next Time: Doing Fraud Detection