MOSCOW, January 10. /TASS/. Researchers from Tyumen State University (TyumSU) discovered a new method of objective translation expertise, according to Maria Kunilovskaya, the principle researcher and associate professor at the Department of English Philology and Translation at TyumSU.
To set the program into motion, the scientists used a text analysis software and a large translation data assembly (corpus-based) put together by scientists from TyumSU and the Higher School of Economics. The developers of this novel analysis tool hope that their program will assist in training more qualified translators. "Our project aims at defining the key features of translation which may form a basis for an automatic linguistic expertise on the quality of a translation, as well as honing the training process for future translators", said Kunilovskaya.
"We highlight the fact that our method is capable of providing expertise only in word choice but it doesn’t estimate the level of its matching with the original text," the researchers warned. The results of such linguistic analysis make it possible to determine the quality of translations of learning materials and to uncover translation plagiarism.
According to Kunilovskaya, it is possible to boost the objectivity of linguistic expertise on the quality of a translation by its automatic comparison with texts of the same genre, translated by professionals. This sort of appraisal detects inevitable mismatches with the original narration, which are peculiar qualities for translations but not for non-translated texts from the same genre written in the same language.
The researchers used the corpus-based methods in linguistics which specializes in the compilation of corpuses - collections of machine-readable exactly-sized texts. The corpuses consist of texts appropriately gathered depending on the goals of the research. For example, the Russian Learner Translator Corpus (RusLTC) consists of students’ English-Russian translations and can be used to identify patterns of beginning translators’ linguistic performance.
In her research, Kunilovskaya evaluated the quality of student translations. She compared the texts of student and professional translations with those from the Russian National Corpus (RNC) using several criteria: sentence length, lexical diversity, information density (number of significant meaningful words), frequency of grammar forms, and frequency of connectors (link words like "for example", "moreover", "therefore").
The result of the text comparison using the big data approach demonstrated the peculiarities of the translation language. For example, both groups of translations appeared to have longer sentences than in the original text, less information density, more function words, less lexical diversity and more link words. Additionally, the research revealed statistically significant differences between professional translations and those made by students which, according to Kunilovskaya, proves that this method can be applied to evaluating a translator’s competence.