查看文章

In this article, aspects regarding the optimisation of machine translation systems via evolutionary computation algorithms are examined. The article focuses on pattern-recognition based machine translation systems that use large monolingual corpora in the target language from which statistical information is extracted. The research reported here uses a specific machine translation as a representative for experimentation. Based on previous studies, SPEA2 is selected as the optimisation method. Issues examined in this article include the effect of population size on the optimisation process and the number of epochs required for the algorithm to settle to near-optimal results. In addition, the effects of different parameters on the translation process are examined, with the aim of reducing the set of system parameters that are actively involved in the optimisation process and thus reducing the optimisation processing time.