DHAKA: Researchers have developed a new computer programme that can
recognize people's emotions based on how they type.
Scientists from the Islamic University of Technology in Bangladesh asked a small group of people to type a block of sample text, and then analysed the keystrokes and characteristics to see if they could identify any of seven different emotional states: joy, fear, anger, sadness, disgust, shame or guilt.
The emotions that the programme recognised with the greatest degree of accuracy were joy (87% of the time) and anger (81% of the time), 'LiveScience' reported.
"If we could build any system that is intelligent enough to interact with humans that involves emotions — that is, it can detect user emotions and change its behaviour accordingly — then using machines could be more effective and friendly," the researchers said.
In the first part of the study, 25 people, ages 15 to 40, retyped two paragraphs from Lewis Carroll's famous novel "Alice's Adventures in Wonderland," and picked one of the emotions that they felt while they were typing: joy, fear, anger, sadness, disgust, shame, guilt, feeling neutral or tired.
The last two options were added in case an individual did not identify with any of the original options.
In the second part of the study, the researchers used software that collected text samples from users, who were prompted every 30 minutes to enter their mental state, choosing from the following possibilities: joy, fear, anger, sadness, disgust, shame, guilt or none of the above.
In this part of the experiment, the text that the users typed did not come from a particular source assigned to them, but was collected during their regular computer use.
The researchers used a special type of software that ran in the background to record all keys pressed by the users as well as the users' press and release times.
The researchers then extracted 19 keystroke attributes from the collected data. Some of the attributes included typing speed in 5 second intervals and the time elapsed between when a particular key was pressed and released.
To analyse the sample texts, the investigators used a standard database of words and sentences that were associated with the seven different emotional states.
The study was published in the journal Behavior & Information Technology.
Scientists from the Islamic University of Technology in Bangladesh asked a small group of people to type a block of sample text, and then analysed the keystrokes and characteristics to see if they could identify any of seven different emotional states: joy, fear, anger, sadness, disgust, shame or guilt.
The emotions that the programme recognised with the greatest degree of accuracy were joy (87% of the time) and anger (81% of the time), 'LiveScience' reported.
"If we could build any system that is intelligent enough to interact with humans that involves emotions — that is, it can detect user emotions and change its behaviour accordingly — then using machines could be more effective and friendly," the researchers said.
In the first part of the study, 25 people, ages 15 to 40, retyped two paragraphs from Lewis Carroll's famous novel "Alice's Adventures in Wonderland," and picked one of the emotions that they felt while they were typing: joy, fear, anger, sadness, disgust, shame, guilt, feeling neutral or tired.
The last two options were added in case an individual did not identify with any of the original options.
In the second part of the study, the researchers used software that collected text samples from users, who were prompted every 30 minutes to enter their mental state, choosing from the following possibilities: joy, fear, anger, sadness, disgust, shame, guilt or none of the above.
In this part of the experiment, the text that the users typed did not come from a particular source assigned to them, but was collected during their regular computer use.
The researchers used a special type of software that ran in the background to record all keys pressed by the users as well as the users' press and release times.
The researchers then extracted 19 keystroke attributes from the collected data. Some of the attributes included typing speed in 5 second intervals and the time elapsed between when a particular key was pressed and released.
To analyse the sample texts, the investigators used a standard database of words and sentences that were associated with the seven different emotional states.
The study was published in the journal Behavior & Information Technology.
No comments:
Post a Comment