Multiscale Analysis of Textual Content Using Eyegaze

Reading of a textual content involves a complex coordination between various parts of brain responsible for visual inputs, language processing, cognitive functions and motor response. In addition, psychological factors like attention and perception play a major role in understanding of the content. Many of these factors get reflected in the behaviour of eye movement, as the content is read. In this paper, we present an approach for analysing a textual content in various scales using eyegaze. The scales include (i) individual fixation characteristics, (ii) saccades and fixations within a line (iii) overall difficulty score of the content. An affordable infrared eye tracking device is used to capture the gaze characteristics in an unobtrusive manner. Two types (easy and difficult) of textual contents are designed for the experiment which are benchmarked using standard readability indices. The fixation characteristics include fixation duration, change in drift direction within a fixation and spatial area of a fixation. Using Analysis of Variance (ANOVA), the former two are found to be statistically significant in distinguishing the two types of contents. Within a line, the spatial distance between fixations and the number of switching between saccades and fixations characterize the flow during a reading where the later is found to be statistically significant. A mixture of two partial sigmoid is used as a mapping function to compute the difficulty score of a content from the significant features. For a given content, the variation of these scores among individual readers, enables us to get deeper insights into their cognitive and psychological aspects.