Normalization in retour simultaneous interpreting: Lexical sophistication in ensemble learning

The study reports significant findings regarding the normalization tendency in retour interpreting or interpreting into the B language. Lexical sophistication indices are used with ensemble learning to classify simultaneous interpreting into the B language (SI into B) and original speeches delivered at the United Nations Security Council (UNSC) conferences. Specifically, SI into B is performed from Chinese to English by interpreters whose first language is Chinese, and the American delegates deliver original speech in English. The ensemble learning achieves an accuracy score of 71%, indicating a discernible distinction between SI into B and original English. This finding further suggests an absence of a significant normalization tendency in SI into B. This can be attributed to the demanding nature of SI tasks and the cognitive load involved in retour interpreting. Differences between SI into B and original English are primarily observed in the indices derived from the BNC spoken corpus. In SI into B, more concrete, imageable, and semantically meaningful words are prevalent, alongside a preference for highly frequent but limitedly distributed words and n-grams. This elucidation can be attributed to the interplay between lexical sophistication indices and the cognitive demands in SI into B.