Software To Understand Your Emotions From The Way You Type On Keyboard
If you thought you could spill a cup of coffee on your keyboard in a furious state of mind after reading an online troll and clean it up later thinking no one will know, well you might be wrong. A team of researchers from Islamic University of Technology in Bangladesh have successfully built a software that can read your emotions on the basis of the pattern of your keyboard strokes & the text. That means the next time you send someone an apology email, they could know what you were thinking while typing it. No points for sarcasm, alright? In the latest research paper, the team from Bangladesh describes how a combination of the text being typed and the keyboard typing patterns of the user can be used by a computer software to predict how the user's emotion with 80% accuracy.
Emotions are one of the most fundamental characteristics in human beings that makes them different than machines. If a system could be built that is intelligent enough to predict the emotions of the human it is interacting with, we would have a smart system that's the next biggest step in affective computing & largely the human-computer interaction model.
So far there have been numerous attempts for determining user emotion through procedures that involve physical sensors attached to user's skin or mechanism for tracking gesture or posture and even voice intonation analysis. All these approaches require expensive and complex hardware which are not only intrusive by difficult to implement. Therefore, in their attempt to find a cheaper solution, these computer science researchers used a standard keyboard as a medium of the user input.
Flow diagram of the team's approach
The team's software collects data (fixed-text using Java-based software & free-text using C# based programs) and extracts attributes from it. To analyse this text, they used the ISEAR dataset for text pattern analysis. VSM or term vector model, an algebraic model was used along with Jaccard similarity as a text pattern classifier. In their experiments, the team asked volunteers to type phrases displayed to them on a screen. They recorded and measured the information on how people typed and asked them about their emotional state (tired, joyful, guilty or disgusted). Later, the software was used to predict how the person was feeling as they typed onto the keyboard. Interestingly enough, the approach showed above 80% accuracies in identifying emotions.
In the future, the team hopes to build a chat application that can assess the emotions of the users based on the text of their conversation. We are imagining a much more sophisticated form of emoticons so that we no more have to add "😔" or "😁" at the end of sentences. It would be a great boon to have the recipient understand the emotional message's tone without having people to ascertain it. In short, a chat app that lets people communicate the most naturally.
What are your thoughts on the application of this kind of a software? Share your comments in replies below.
Source: #-Link-Snipped-#
Emotions are one of the most fundamental characteristics in human beings that makes them different than machines. If a system could be built that is intelligent enough to predict the emotions of the human it is interacting with, we would have a smart system that's the next biggest step in affective computing & largely the human-computer interaction model.
So far there have been numerous attempts for determining user emotion through procedures that involve physical sensors attached to user's skin or mechanism for tracking gesture or posture and even voice intonation analysis. All these approaches require expensive and complex hardware which are not only intrusive by difficult to implement. Therefore, in their attempt to find a cheaper solution, these computer science researchers used a standard keyboard as a medium of the user input.
Flow diagram of the team's approach
The team's software collects data (fixed-text using Java-based software & free-text using C# based programs) and extracts attributes from it. To analyse this text, they used the ISEAR dataset for text pattern analysis. VSM or term vector model, an algebraic model was used along with Jaccard similarity as a text pattern classifier. In their experiments, the team asked volunteers to type phrases displayed to them on a screen. They recorded and measured the information on how people typed and asked them about their emotional state (tired, joyful, guilty or disgusted). Later, the software was used to predict how the person was feeling as they typed onto the keyboard. Interestingly enough, the approach showed above 80% accuracies in identifying emotions.
In the future, the team hopes to build a chat application that can assess the emotions of the users based on the text of their conversation. We are imagining a much more sophisticated form of emoticons so that we no more have to add "😔" or "😁" at the end of sentences. It would be a great boon to have the recipient understand the emotional message's tone without having people to ascertain it. In short, a chat app that lets people communicate the most naturally.
What are your thoughts on the application of this kind of a software? Share your comments in replies below.
Source: #-Link-Snipped-#
0
