Eye Tracking as an Automatic MT Evaluation Technique

OPEN ARCHIVE

Union Jack
Dannebrog

Eye Tracking as an Automatic MT Evaluation Technique

Vis flere oplysninger

Titel: Eye Tracking as an Automatic MT Evaluation Technique
Forfatter: Carl, Michael; Doherty, Stephen; O’Brien, Sharon
Resume: Eye tracking has been used successfully as a technique for measuring cognitive load in reading, psycholinguistics, writing, language acquisition etc for some time now. Its application as a technique for automatically measuring the reading ease of MT output has not yet, to our knowledge, been tested. We report here on a preliminary study testing the use and validity of an eye tracking methodology as a means of semi- and/or automatically evaluating machine translation output. 50 French machine translated sentences, 25 rated as excellent and 25 rated as poor in an earlier human evaluation, were selected. 10 native speakers of French were instructed to read the MT sentences for comprehensibility. Their eye gaze data were recorded non-invasively using a Tobii 1750 eye tracker. The average gaze time and fixation count were found to be higher for the “bad” sentences, while average fixation duration and pupil dilations were not found to be substantially different between output rated as good or bad. Comparisons between BLEU scores and eye gaze data were also made and found to correlate well with gaze time and fixation count, and to a lesser extent with pupil dilation and fixation duration. We conclude that the eye tracking data, in particular gaze time and fixation count, correlate reasonably well with human evaluation of MT output but fixation duration and pupil dilation may be less reliable indicators of reading difficulty for MT output. We also conclude that eye tracking has promise as an automatic MT Evaluation technique.
URI: http://hdl.handle.net/10398/8045
Dato: 2010-04-15

Creative Commons License This work is licensed under a Creative Commons License.

Filer Størrelse Format Vis
SubmissionforMT_dohertyobriencarl.pdf 220.9Kb PDF Vis/Åbn Preprint version

Dette dokument findes i følgende samling(er)

Vis flere oplysninger