Date of Award
Bachelor of Science Honours
Faculty of Computing, Health and Science
Translation costs restrict the preparation of medical survey and other questionnaires for migrant communities in Western Australia. This restriction is compounded by a lack of affordable and accurate machine translation mechanisms. This research investigated and evaluated combined strategies intended to provide an efficacious and affordable machine translator by: • using an interlingua or pivot-language that requires less resources for its construction than contemporary systems and has the additional benefit of significant error reduction; and • defining smaller lexical environments to restrict data, thereby reducing the complexity of translation rules and enhancing correct semantic transfer between natural languages. This research focussed on producing a prototype machine translation mechanism that would accept questionnaire texts as discrete questions and suggested answers from which a respondent may select. The prototype was designed to accept non-ambiguous English as the source language, translate it to a pivot-language or interlingua, Esperanto, and thence to a selected target language, French. Subsequently, a reverse path of translation from the target language back to the source language enabled validation of minimal or zero change in both syntax and semantics of the original input. Jade, an object-oriented (00) database application, hosting the relationship between the natural languages and the interlingua, was used to facilitate the accurate transfer of meaning between the natural languages. Translation, interpretation and validation of sample texts was undertaken by linguists qualified in English, French and Esperanto. Translation output from the prototype model was compared, again with assistance from linguists, with a 'control' model, the SYSTRAN On-Line Translator, a more traditional transfer translation product. Successful completion of this research constitutes a step towards an increased availability of low cost machine translation to assist in the development of reliable and efficient survey translation systems for use in specific user environments. These environments include, but arc not exclusive to, medical, hospital and Australian indigenous-contact environments.
Boddington, R. (2004). Evaluation of an Esperanto-Based Interlingua Multilingual Survey Form Machine Translation Mechanism Incorporating a Sublanguage Translation Methodolgy. Retrieved from https://ro.ecu.edu.au/theses_hons/954