IBM’s deep question-answering system, Watson, gained a great deal of fame after its appearance on the television quiz show “Jeopardy” in February. And now, IBM is trying to find ways to demonstrate the technology in compelling ways while getting the next generation of developers—today’s students—excited about it.
Already, students have suggested applications for healthcare data, travel search engine data and government database information, according to Jim Spohrer, director of university programs at IBM.
IBM recently joined with Carnegie Mellon University and the University of Pittsburgh to host a symposium for students to brainstorm about the future of the technology platform.
Carnegie Mellon, led by Eric Nyberg (professor at the language technologies institute at CMU’s school of computer science), assisted IBM in creating the Watson platform. Spohrer said that CMU’s involvement contributed to IBM’s decision to work with students at the two schools.
“Any industry that has had an information explosion, like healthcare, can use Watson,” said Spohrer. “Watson is going to be valuable [in industries] where the half-life of knowledge is getting shorter and shorter. Government services—taxes, healthcare benefits, retirement benefits—will also benefit.”
Spohrer also said social media applications, such as travel websites, could utilize the technology, as would anything that needs a deep question-and-answering service to understand and interpret data.
Nico Schlaefer, a doctoral student at Carnegie Mellon University, said that Watson is a contained system with all the data residing in the platform. So, for instances where connectivity might be an issue, Watson can still perform at a high level, he said.
This, he said, is a valuable tool for industries looking to create an application that can provide answers independently of the Internet’s resources. An additional benefit is improved performance, as applications with their own data stored within the code work more quickly than those that have to access a remote database.
Schlaefer said the travel industry could benefit most from the source expansion algorithm contributed by CMU in its applications, but generating self-contained data can be done for any topic. He said that for medical questions, one would have to add medical journals and perhaps scientific articles, and then allow the platform to search for additional resources.
Schlaefer worked on Nyberg’s team, and he has also interned at IBM for the past three summers. His doctoral thesis focuses on the source expansion algorithm that allows Watson to be a contained system capable of answering questions on a wide variety of topics.
Data is imported to Watson in two ways: manually, or by a search of the Web. Schlaefer explained that the source expansion algorithm takes the data added manually and searches for and collects similar material from the Web, creating a massive, unstructured database within Watson. The system then relies on statistical data to determine if the information found on the Web is accurate.
CMU also contributed an answer-scoring algorithm that improved Watson’s ability to recognize when a possible answer was likely to be correct.
“Watson could not communicate with the Web for the ‘Jeopardy’ contest because the other contestants couldn’t, so all the data had to be self-contained,” said Schlaefer. “We gathered encyclopedias and other data resources manually and then added this algorithm” to search past “Jeopardy” questions on the Web to see how accurate Watson’s answers would have to be to win the game.
Schlaefer added that the Watson platform’s ability to take natural English questions is what differs its search capabilities from the likes of search giants, such as Google.
“Google searches unstructured data [the Web] and returns a wide variety of results, based on keywords. Watson searches natural English questions and returns precise and relevant results,” Schlaefer said.
Shortly after Watson’s triumph on “Jeopardy” (the computer defeated two of the greatest human champions in the show’s history), IBM announced that doctors at the Columbia University Medical Center and the University of Maryland School of Medicine will use the technology to assist in examining medical information and diagnoses, along with other medical software applications.
Nuance Communications has also entered a partnership with IBM to apply Watson to its healthcare software applications.