A Comparative Study Of Linguistic Quantitative Features In Human And Google Translations Of Fiction And Social Science Texts

The need to evaluate machine translation performance has become increasingly urgent with the rapid development and growing popularity of machine translation over the past decade.In response to this demand,scholars have made efforts to develop methods to assess the strengths and weaknesses of machine translation.However,it has been noted that there is still a gap in research focused on specific text types and language pairs.This study employs the principle of translation quality assessment and quantitative linguistics to compare human-translated texts and Google-translated texts from two distinct genres——fiction and social science.The language pair under consideration is Chinese to English.To facilitate the study,the researcher built a self-made corpus comprising of 30 fiction texts and 30 social science texts in Chinese,along with their corresponding Googletranslated and human-translated versions.Using language processing tools,the researcher computed various linguistic indicators and conducted Wilcoxon signed-rank tests with Bonferroni correction to identify significant differences between the translations.The study identified significant differences between Google translation and human translation in 148 indicators for fiction texts and 37 indicators for social science texts.The results showed that Google Translate performs better in translating social science texts than fiction texts in more indicators,with more grammatical mistakes found in its translation of fiction texts.The study discovered that Google Translate exhibits conservative word choices in translating both text types,relying heavily on high-frequency and repeated words,while failing to utilize strong and meaningful verbs.In its translation of fiction texts,it indiscriminately selects tenses,and in its translation of social science texts,it repeated verbs and adverbs excessively.At the syntactic level,it tends to translate literally,limiting its flexibility in using noun phrases and handling sentences with a passive voice in the fiction group.In the social science group,it fails to use gerunds effectively to conform to the static feature of English.At the discoursal level,Google Translate struggles with recognizing logical relationships in Chinese texts to enhance deep cohesion,and with using pronouns to improve coherence and avoid repetition.In its translation of social science texts,it needs to expand translation units to identify the communicative function.In conclusion,Google Translate needs to improve its vocabulary choices,avoid literal translations,conform to English writing conventions,and strengthen deep cohesion of texts.It is hoped that the findings can provide insights for the improvement of Google Translate and for future translation teaching by guiding human translators in AI-supported translation.