Cross-lingual Terminology Extraction for Translation Quality Estimation

Cross-lingual Terminology Extraction for Translation Quality Estimation⋆ Yu Yuan† ∗, Yuze Gao‡, Yue Zhang‡, Serge Sharoff∗ † School of Languages & Cultures, Nanjing University of Information Science & Technology, 210044, China ‡ Information Systems Technology and Design, Singapore University of Technology and Design, 487372 ∗ Centre for Translation Studies, University of Leeds, LS2 9JT, United Kingdom hittle.yuan@gmail.com, {yuze gao, yue zhang}@sutd.edu.sg s.sharoff@leeds.ac.uk Abstract We explore ways of identifying terms from monolingual texts and integrate them into investigating the contribution of terminology to translation quality.The researchers proposed a supervised learning method using common statistical measures for termhood and unithood as features to train classifiers for identifying terms in cross-domain and cross-language settings. On its basis, sequences of words from source texts (STs) and target texts (TTs) are aligned naively through a fuzzy matching mechanism for identifying the correctly translated term equivalents in student translations. Correlation analyses further show that normalized term occurrences in translations have weak linear relationship with translation quality in term of usefulness/transfer, terminology/style, idiomatic writing and target mechanics and near- and above-strong relationship with the overall translation quality. This method has demonstrated some reliability in automatically identifying terms in human translations. However, drawbacks in handling low frequency terms and term variations shall be dealt in the future. Keywords: Bilingual terminology, translation quality, supervised learning, correlation analysis 1. Introduction Terminology helps translators organize their domain knowledge, and provides them means (usually terms in var- ious lexical units) to express subject knowledge adequately. Translation scholars and practitioners maintain that termi- nology correctness is associated with the quality of trans- lation (and interpretation) (Hartley et al., 2004; Xu and Sharoff, 2014; Kim et al., 2015; Brunette, 2000; Karoubi, 2016). The acknowledgement of the contribution of terminology to translation quality is also echoed by the translation industry and users (Secar˘a, 2005; Lommel et al., 2014; Warburton, 2013). Accurately reproducing the content of the original and using appropriate terminology has become the official assessment criteria of some famous in-use translation-error- based evaluation schemes. For instance, the MeLLANGE project (Secar˘a, 2005) defines more than six terminology errors1, and the Multidimensional Quality Metrics lists ter- minology as one of the eight major dimensions, which is subdivided into three children issue types (term inconsis- tency, termbase2, and terminology domain3) (Lommel et al., 2014). From a user’s expectation perspective, appro- priate terminological use is also viewed as one of the im- portant quality parameters. For the purpose of marketing, companies will localize the manuals that accompany their products. Localization cannot be done at the expense of quality to endanger the customer satisfaction. Their dissat- isfaction will lead to more potential damaging losses in rev- This work is done when the first author works as a research fellow at SUTD. 1The main terminological errors are incorrect terminology, false cognate, term translated by non-term, inconsistent with glos- sary, inconsistent within target text (TT), inappropriate colloca- tion, and user-defined errors. 2a term is translated is translated with a term nonconforming to the specification. 3a term is translated with a term …

Content truncated to 500 words