Dynamic Programming for Re-Mapping Noisy Fixations in Translation Tasks

Eyetrackers which allow for free head movement are in many cases imprecise to the extent that reading pattern become heavily distorted. As a result, a "naïve" gazeto-symbol mapping approach wrongly maps also the drifted center of the xation on the symbol directly below it. In this talk, I extend this "naïve" gaze-to-word mapping by introducing background knowledge about the gazing task. In a first step, the sequence of "naïve" gaze-to-symbol mappings is projected into a lattice of several possible gaze locations above and below the current fixation on the text. In a second step a dynamic programming algorithm applies a number of heuristics to nd the best path through the lattice, based on the likely distance in characters and in pixels between successive fixations, so as to smooth the gazing path according to observations reported in the established literature.
I present several types of drifting problems and noise that we observe when using eyetrackers to record and analyse gazing behavior of translators. I describe the mapping algorithm and show how it rectifies and re-aligns the noisy gaze data to provide a more intelligible gaze path. I show examples of re-mapped fixations which better explain translator's activity than naïve gaze-to-symbol mapping.