IBM’s Watson Could Make a Knowledgeable Tour Guide
IBM researchers are exploring Watson’s abilities to answer museumgoers’ questions.
Machine learning has the potential to help us sort through data and find answers to all kinds of questions.
IBM’s Watson, the machine-learning computer that won Jeopardy! in 2011 and has found work searching medical and scientific data for insights, could soon have yet another job: museum tour guide.
A group of researchers at IBM Research India used Watson as part of an Android app called Usher that supplies information about nearby artwork in a museum and answers questions that you ask about it. It’s an example of the way IBM hopes programmers will make use of the capabilities of Watson in all sorts of apps. The research was presented last week at a conference on intelligent user interfaces in Atlanta.
The app analyzes accelerometer data from the phone to tell whether the user is walking, roaming around an area, or standing still; and it analyzes accelerometer, gyroscope, and compass data to figure out where the user is looking. That way, the app can tell you things like, “On your left is van Gogh’s Sunflowers.”
When the person stopped in front of a painting, researchers could type or speak a question about an artwork being viewed, such as “Who painted this?” or “What is the meaning of this painting?” or “What is the medium of this painting?” The person’s location and the question are then sent to a version of Watson—in the case of the simulation, a version trained to answer general questions.
Usher isn’t publicly available yet, and has been tried out only with some demo content that researchers created to go along with some paintings. But Shubham Toshniwal, a coauthor of the paper who works in the Watson group at IBM Research India in New Delhi, says it has been received “pretty well” internally at IBM, and he can imagine it helping people learn more about their surroundings in a variety of indoor spaces, ranging from historical buildings to offices.
In the years since Watson’s star turn on TV, IBM has poured money into the research effort behind it in hopes that it will yield software and hardware that can sift through mountains of data to answer complicated questions (see “Does Watson Know the Answer to IBM’s Woes?”).
With Usher, the IBM researchers used software to simulate how a person would browse a museum in which different artworks and areas were assigned IDs that corresponded with data (such as the names of paintings) stored in a database. The person’s location, tracked indoors by measuring the strength of available Wi-Fi signals in relation to the user’s smartphone, would help Usher determine which artworks they were near.
Toshniwal says the app can find your friends who are also in the museum by connecting accounts from social networks like Facebook.
If researchers continue working on Usher, one big issue is the need to improve indoor location tracking. As the research notes, while Wi-Fi signals can give basic data about where a person is standing, it’s not precise.
A potential solution may lie in the use of proximity sensors like Apple’s iBeacon technology, which uses low-energy Bluetooth to determine the proximity of nearby users’ iPhones. A number of museums are already trying it out.
Keep up with the latest in AI at EmTech MIT.
Discover where tech, business, and culture converge.
September 11-14, 2018
MIT Media Lab