Eye Reader
‘Machine Learning as a Tool for Interactive Products' class
School of Computer Science and Engineering at the Hebrew University of Jerusalem.
Guided by Dr. Fink Michael. 2014.
What if the computer could know when the reader doesn't understand and provide translation or reading assistance in real time?
EyeReader is a research project, aspiring to create an improved reading experience using machine learning. It can help in learning a new language, reading a complicated text, or simply understanding menus while traveling abroad. In order to do so I have used an eye tracker. Eye tracker is a sensor that follows the eyes movement, therefore allowing the computer to know exactly where the eyes are focused. The blue circle is the average position of the reader's pupils.
The information received from the eye tracker is then used by a machine learning algorithm called SVM - Support Vector Machine - to binary classify each word to known and unknown.
In addition to the eyes position, the algorithm uses several other parameters, such as the word’s length, location within the entire text, frequency in the English language, and the time spent on the word (normalized according to number of appearances in the text). After the initial training, the algorithm managed to identify correctly words as unknown, in real time.
Text from Hamlet
Act 3, scene 1, page 4, lines 71-83.