Lexical ambiguity in machine translation and its impact on the evaluation of output by users

Ambiguity has often been described as a „key bottleneck for progress in Machine Translation“ (Dale, Moisl and Somers 2000). While several studies have been centered on the description of ambiguity as a linguistic and philosophical phenomenon, there has been no study, to the best of our knowledge, that has measured its impact on machine translation output. This quantitative case study highlights the influence of ambiguity on machine translation’s output quality in the judgement of users, thus providing concrete data to measure this impact. Our experiments also shed light on the fact that no absolute correlation exists between fluency- and adequacy scores. Lastly, this dissertation also includes a comprehensive survey of different forms of ambiguities and ambiguity resolution techniques in MT.