Investigating the Influence of Bilingual MWU on Trainee Translation Quality

Investigating the Influence of Bilingual Multiword Units on Trainee Translation Quality Yu Yuan† ‡, Serge Sharoff† ‡ School of Languages & Cultures, Nanjing University of Information Science & Technology, 210044, China † Centre for Translation Studies, University of Leeds, LS2 9JT, United Kingdom hittle.yuan@gmail.com, s.sharoff@leeds.ac.uk Abstract We applied a method for automatic extraction of bilingual multiword units (BMWUs) from a parallel corpus in order to investigate their contribution to translation quality in terms of adequacy and fluency. Our statistical analysis is based on generalized additive modelling. It has been shown that normalised BMWU ratios can be useful for estimating human translation quality. The normalized alignment ratios for BMWUs longer than two words have the greatest impact on measuring translation quality. It is also found that the alignment ratio for longer BMWUs is statistically closer to adequacy than to fluency. Keywords: Bilingual multiword unit, normalised alignment ratio, translation quality estimation, trainee translators 1. Introduction The overall goal of this study is to investigate auto- matic quality assessment for translators, especially for the trainees. Many problems related to quality stem from inap- propriate translation of MWUs, such as idioms (e.g. bark- ing at the wrong tree, add insult to injury, etc.), terms (e.g. recursive function, closed captioning, etc.), phrasal verbs (e.g. fall out, give up, etc.), or named entities (e.g. Vice- Chancellor Sir Timothy Michael Martin O’Shea, New York City, etc.). Baker (2011, pp:72-73) identified four types of errors when translating idiomatic expressions: ‘no equiv- alent’, ‘similar counterpart but different context’, ‘idiom play’, ‘different conventions of using idioms’. A trainee translator in our dataset produced an unacceptable calque of as brave as a lion into Chinese 勇猛如狮, while the correct idiom should have been 勇猛如虎 ‘as brave as a tiger’. This erroneous treatment is what Baker terms ‘dif- ferent conventions of using idioms’. Based on observations of source language (SL) idioms and their Persian translation, researhers have found that ex- plicit loss,implicit loss, modified loss and complete loss are common resultant categories of cultural losses (Zebardast and AbuSaeedi, 2015; Zebardast, 2015). Translation of MWUs is problematic because it is associated with cultural issues (Min, 2007), systematic language variation (Wang and Nian, 2004) and the translator’s failure to decipher the meanings of MWUs in question (Abu-Ssaydeh, 2004). Therefore, this paper assesses the contribution of translat- ing MWUs to the overall translation quality, particularly for trainee translations. In this paper we investigate: 1. how can we extract BMWUs for trainee translation quality evaluation? 2. how is the use of BMWUs related to the quality of trainee translations? 3. How do BMWUs of different lengths contribute to the trainee translation quality? The term BMWUs refers to recurrent sequences of words that are translations to each other in aligned bilingual texts. In other studies they are also known as (bilingual) phrase alignments, aligned phrases units (sequences), bilin- gual alignments, bilingual phraseological units, bilingual phraseology, etc. In this paper, we use the term bilingual multiword units (BMWUs) hereinafter. We start with the automatic acquisition of BMWUs from …

Content truncated to 500 words