Analysing terminology translation errors in statistical and neural machine translation
- PDF / 1,319,478 Bytes
- 47 Pages / 439.37 x 666.142 pts Page_size
- 48 Downloads / 314 Views
Analysing terminology translation errors in statistical and neural machine translation Rejwanul Haque1 · Mohammed Hasanuzzaman1 · Andy Way1 Received: 17 April 2019 / Accepted: 23 July 2020 © Springer Nature B.V. 2020
Abstract Terminology translation plays a critical role in domain-specific machine translation (MT). Phrase-based statistical MT (PB-SMT) has been the dominant approach to MT for the past 30 years, both in academia and industry. Neural MT (NMT), an end-to-end learning approach to MT, is steadily taking the place of PB-SMT. In this paper, we conduct comparative qualitative evaluation and comprehensive error analysis on terminology translation in PB-SMT and NMT in two translation directions: English-to-Hindi and Hindi-to-English. To the best of our knowledge, there is no gold standard available for evaluating terminology translation quality in MT. For this reason we select an evaluation test set from a legal domain corpus and create a gold standard for evaluating terminology translation in MT. We also propose an error typology taking the terminology translation errors in MT into consideration. We translate sentences of the test set with our MT systems and terminology translations are manually classified as per the error typology. We evaluate the MT system’s performance on terminology translation, and demonstrate our findings, unraveling strengths, weaknesses, and similarities of PB-SMT and NMT in the area of term translation. Keywords Terminology translation · Machine translation · Phrase-based statistical machine translation · Neural machine translation
* Rejwanul Haque [email protected] Mohammed Hasanuzzaman [email protected] Andy Way [email protected] 1
ADAPT Centre, School of Computing, Dublin City University, Dublin, Ireland
13
Vol.:(0123456789)
R. Haque et al.
1 Introduction Terms are productive in nature, and new terms are being created all the time. A term could have multiple meanings depending on the context in which it appears. For example, the words “terminal” (‘a bus terminal’ or ‘terminal disease’ or ‘computer terminal’) and “play” (‘play music’ or ‘plug and play’ or ‘play football’ or ‘a play’) could have very different meanings depending on the context in which they appear. A polysemous term (e.g. terminal) could have many translation equivalents in a target language. For example, the English word ‘charge’ has more than twenty target equivalents in Hindi (e.g. ‘dam’ for ‘value’, ‘bhar’ for ‘load’, ‘bojh’ for ‘burden’). When encountering a judicial document, the translation of “charge” has to be the particular Hindi word: ‘aarop’. The target translation could lose its meaning if the term translation and domain knowledge are not taken into account. Accordingly, the preservation of domain knowledge from source to target is pivotal in any translation workflow (TW), and this is one of the customer’s primary concerns in the translation industry. Naturally, translation service providers (TSPs) who use MT in their production expect translations t
Data Loading...
