Improving Subword-level Translation Quality

Author(s):  
Anoop Kunchukuttan ◽  
Pushpak Bhattacharyya
Keyword(s):  
Electronics ◽  
2021 ◽  
Vol 10 (13) ◽  
pp. 1589
Author(s):  
Yongkeun Hwang ◽  
Yanghoon Kim ◽  
Kyomin Jung

Neural machine translation (NMT) is one of the text generation tasks which has achieved significant improvement with the rise of deep neural networks. However, language-specific problems such as handling the translation of honorifics received little attention. In this paper, we propose a context-aware NMT to promote translation improvements of Korean honorifics. By exploiting the information such as the relationship between speakers from the surrounding sentences, our proposed model effectively manages the use of honorific expressions. Specifically, we utilize a novel encoder architecture that can represent the contextual information of the given input sentences. Furthermore, a context-aware post-editing (CAPE) technique is adopted to refine a set of inconsistent sentence-level honorific translations. To demonstrate the efficacy of the proposed method, honorific-labeled test data is required. Thus, we also design a heuristic that labels Korean sentences to distinguish between honorific and non-honorific styles. Experimental results show that our proposed method outperforms sentence-level NMT baselines both in overall translation quality and honorific translations.


2021 ◽  
Vol 54 (2) ◽  
pp. 1-36
Author(s):  
Sameen Maruf ◽  
Fahimeh Saleh ◽  
Gholamreza Haffari

Machine translation (MT) is an important task in natural language processing (NLP), as it automates the translation process and reduces the reliance on human translators. With the resurgence of neural networks, the translation quality surpasses that of the translations obtained using statistical techniques for most language-pairs. Up until a few years ago, almost all of the neural translation models translated sentences independently , without incorporating the wider document-context and inter-dependencies among the sentences. The aim of this survey article is to highlight the major works that have been undertaken in the space of document-level machine translation after the neural revolution, so researchers can recognize the current state and future directions of this field. We provide an organization of the literature based on novelties in modelling and architectures as well as training and decoding strategies. In addition, we cover evaluation strategies that have been introduced to account for the improvements in document MT, including automatic metrics and discourse-targeted test sets. We conclude by presenting possible avenues for future exploration in this research field.


Babel ◽  
2015 ◽  
Vol 61 (2) ◽  
pp. 283-303 ◽  
Author(s):  
Lily Lim ◽  
Kwok Ying Loi

Slogans play an important role in conveying information to targeted audiences, and the translation of slogans tends to be studied under the rubric of public-notice translation. Previous research mainly uses researchers’ expertise to evaluate the quality of this type of translation. Yet, little is known about what the targeted readers think about the translation, although their opinions present key information that helps to determine whether the translation has achieved the intended effect. This paper elicits and systematically analyzes readers’ responses. We investigate the case of Macao, a rapidly growing economy where the demand for English translation has markedly increased in recent decades. Public administration bodies in Macao have commissioned Chinese-to-English translation in varied areas such as tourism, social security and welfare, cultural and sports events. We sampled ten translated slogans that were used in the public sector, and administered survey questionnaires (n=130) to both source-text and target-text readers. The two groups of readers’ evaluations, based on the criteria of fluency, conciseness, persuasiveness and mnemonic effect, reveal that the translations are perceived significantly less favorably than the originals are. Readers most strongly disliked word-for-word translations, and pointed out numerous problems with the translations such as ungrammaticality, inappropriate word use, lack of appeal, and unintelligibility due to insufficient background knowledge. This research demonstrates the tangible value of using readers’ responses to evaluate translation quality. It also has implications for translator training, and recommends that public authorities should institute a rigorous quality assurance system.


Target ◽  
2014 ◽  
Vol 26 (1) ◽  
pp. 147-150
Author(s):  
Sharon O’Brien
Keyword(s):  

2013 ◽  
Vol 27 (3-4) ◽  
pp. 281-301 ◽  
Author(s):  
Jesús González-Rubio ◽  
J. Ramón Navarro-Cerdán ◽  
Francisco Casacuberta

2021 ◽  
pp. 1-10
Author(s):  
Zhiqiang Yu ◽  
Yuxin Huang ◽  
Junjun Guo

It has been shown that the performance of neural machine translation (NMT) drops starkly in low-resource conditions. Thai-Lao is a typical low-resource language pair of tiny parallel corpus, leading to suboptimal NMT performance on it. However, Thai and Lao have considerable similarities in linguistic morphology and have bilingual lexicon which is relatively easy to obtain. To use this feature, we first build a bilingual similarity lexicon composed of pairs of similar words. Then we propose a novel NMT architecture to leverage the similarity between Thai and Lao. Specifically, besides the prevailing sentence encoder, we introduce an extra similarity lexicon encoder into the conventional encoder-decoder architecture, by which the semantic information carried by the similarity lexicon can be represented. We further provide a simple mechanism in the decoder to balance the information representations delivered from the input sentence and the similarity lexicon. Our approach can fully exploit linguistic similarity carried by the similarity lexicon to improve translation quality. Experimental results demonstrate that our approach achieves significant improvements over the state-of-the-art Transformer baseline system and previous similar works.


Target ◽  
2021 ◽  
Author(s):  
Silvia Parra-Galiano

Abstract This article proposes a hierarchy of translator and reviser competences in prototypical scenarios in legal translation with a view to determining the most appropriate revision foci to ensure translation quality. Built on a prior characterisation of the most common professional translator profiles in legal translation, the proposal for a hierarchy of competences derives from two premises: (1) The professional profile of those who translate and revise legal documents is very diverse in terms of competence and qualifications (training and experience), and (2) translation competence and specialist knowledge in legal fields (i.e., domain competence) are fundamental when revising to guarantee the quality of legal translations. The proposal is framed by quality assurance in legal translation through a revision process based on (1) the coherent management of the work of the translators and revisers involved in the translation project, and (2) the appropriate methodology for revision applied to legal translation by adapting the revision mode’s focus to ensure its effectiveness. Six common scenarios are identified in light of the translators’ profiles, for which revisers’ profiles are then proposed in order to detect any legal translation competence deficiencies among translators, and thus ensure quality.


Sign in / Sign up

Export Citation Format

Share Document