| The computer is beginning to facilitate the analysis of texts, and is allowing us to view written language in new and interesting ways. The results of such analyses might have important pedagogical implications for the second language classroom. This study specifically examines the output of two Machine Translation (MT) software programs searching for patterns that may be of interest to second language educators. If studying student errors in the 1960s and 70s greatly contributed to our understanding of various aspects second language acquisition, what might be gained by studying the error patterns of machine translation programs designed by expert linguists? Since both must be ‘taught’ to overcome many of the same linguistic hurdles whenever they communicate in L2, it seems logical that studying the error patterns of machine translation systems might prove equally beneficial.; The findings of this study revealed a number of interesting results. Lexical errors greatly outnumbered grammatical errors in the evaluated MT texts. This was especially true of prepositions which accounted for a large number of lexical errors due to their polysemic nature. Grammatically, verbs presented the biggest grammatical challenge because the verb tenses of one language do not neatly correspond with those of another language. Reflexive verbs were particularly problematic and consequently were mistranslated almost everywhere they occurred in text.; The source or first language (L1) seemed to have a strong influence on the errors produced in the target or second language (L2). Finally, the study shows that people use more fixed or patterned language than they may consciously realize. |
