Are lexical and syntactic simplification and semantic compression techniques effective in machine translation? A comparison of cognitive processing and performance effects in human and machine evaluators

When:6 Sep 2016, 3pm - 4:30pm
Venue:Quadrangle G031 (map Ref E15)
Who:Dr Stephen Doherty (UNSW)
Dr Stephen Doherty

Interpreting & Translation seminar


Lexical and syntactic simplification and semantic compression techniques have consistently been shown to be effective in monolingual contexts, resulting in increased readability, comprehension, and recall across a variety of tasks. Applied to bilingual contexts, these techniques have also been shown to increase the translatability of texts, from one language into another, by human translators.

This paper describes a series of experiments from a project that aimed to test the efficacy of these techniques in machine translation across five language pairs (English ? French, English ? Spanish, English ? German, English ? Japanese, and English ? Chinese). I show how these techniques are largely effective in all language pairs as measured by a novel triangulation of human and machine evaluation methods. From the development of this methodology and the project’s findings, I identify a point of divergence between human and machine evaluators, where the former are more sensitive to effects at the syntactic level and the latter at the lexical level.

I conclude by accounting for the inconsistent correlations reported between human and machine evaluation results and draw upon more recent work that seeks to address this ongoing issue. Finally, I describe a prototype system from this project which uses eye tracking for semi-automatic, mono- and bilingual linguistic evaluations, and close by providing examples of its industry and public sector application.

About Stephen Doherty

Stephen Doherty, BA (Hons.), HDip, PhD, MBPsS, is a Senior Lecturer and Program Co-Convenor in the School of Humanities & Languages, Faculty of Arts & Social Sciences, at UNSW. His research investigates the cognitive aspects of language processing and language technologies using eye tracking, psychometrics, and electroencephalography. Webpage:

Searleâ??s Chinese Room Experiment,

Image Credit: Searle’s Chinese Room Experiment,

View Past Events