Go ahead and try to hide how you're feeling. Your keystrokes could rat you out.
Researchers in Bangladesh have designed a computer program that analyzes typing and text patterns to identify emotional states. While both methods have been used before to detect computer users' emotions, the researchers combined them, for the first time, they say, and saw promising results -- software that nails mood as much as 87 percent of the time.
A.F.M. Nazmul Haque Nahin and his colleagues at the Islamic University of Technology detail their work in the latest issue of the journal Behavior and Information Technology. They say the findings could be significant to the development of emotionally aware computer systems as their approach relies on less expensive, and less intrusive, methods than tools like voice analysis, facial sensors, thermal imaging, and gesture tracking.
"Combining both keyboard dynamics with text pattern analysis can be more effective to detect user emotion...however, no work has been done on such a combined approach," the scientists say in their paper. "This combined analysis does not require any extra hardware or specialized tools."
The team asked volunteers to note their emotional states after typing passages of fixed text, and while writing regular, or "free," text. The researchers then culled data about keystroke attributes associated with seven states (joy, fear, anger, sadness, disgust, shame, and guilt). To help them analyze sample texts, they used a standard database of words and sentences associated with the same emotional states.
Past research into the overlap of keystrokes and emotion has examined actions like typing speed and use of the backspace key. The scientists from Bangladesh considered these and other factors, such as the number of characters typed every five seconds and how much time people lingered on a given key.
But who other than your cubicle mate cares if you're pounding away at your keyboard in a fit of rage? And why does your computer need to know if you're feeling elated or afraid?
"Computer systems that can detect user emotion can do a lot better than the present systems in gaming, online teaching, text processing, video and image processing, user authentication, and so many other areas where user emotional state is crucial," say the researchers, who describe some inefficiencies in their work and plan to continue refining it.
Imagine a gaming app that could sense a player's emotional state and change the graphics, volume, and music selection accordingly. Or an online course that could swap to a simpler interface if a student showed continued frustration. Or maybe one day, a tool that knows how CNET readers are really feeling when they post a comment.