A Grammatical Analysis on Machine Translation Errors

Abstract Since entering the new era, the translation model has gradually changed with the widespread application of machine translation technology and the rapid development of a translation industry. The mismatch between the demand of employers and the talents trained by universities has become a major problem facing the translation major nowadays. To this end, we should attach more importance to the readjustment of the existent curriculum; students’ practical ability in translation; grasp of the skill of detecting and correcting machine translation errors; combination of translation and relevant professional knowledge.

Download Full-text

Improving Neural Machine Translation by Filtering Synthetic Parallel Data

Entropy ◽

10.3390/e21121213 ◽

2019 ◽

Vol 21 (12) ◽

pp. 1213

Author(s):

Guanghao Xu ◽

Youngjoong Ko ◽

Jungyun Seo

Keyword(s):

Machine Translation ◽

Synthetic Data ◽

Similarity Score ◽

Target Language ◽

Neural Machine Translation ◽

Novel Approach ◽

Parallel Data ◽

Back Translation ◽

Translation Errors ◽

Training State

Synthetic data has been shown to be effective in training state-of-the-art neural machine translation (NMT) systems. Because the synthetic data is often generated by back-translating monolingual data from the target language into the source language, it potentially contains a lot of noise—weakly paired sentences or translation errors. In this paper, we propose a novel approach to filter this noise from synthetic data. For each sentence pair of the synthetic data, we compute a semantic similarity score using bilingual word embeddings. By selecting sentence pairs according to these scores, we obtain better synthetic parallel data. Experimental results on the IWSLT 2017 Korean→English translation task show that despite using much less data, our method outperforms the baseline NMT system with back-translation by up to 0.72 and 0.62 Bleu points for tst2016 and tst2017, respectively.

Download Full-text

Hva er viktig for forståelse? Om maskinoversetting fra nordsamisk

Oslo Studies in Language ◽

10.5617/osla.8514 ◽

2021 ◽

Vol 11 (2) ◽

pp. 489-501

Author(s):

Trond Trosterud ◽

Lene Antonsen

Keyword(s):

Machine Translation ◽

Text Comprehension ◽

The Other ◽

Translation System ◽

Rule Based ◽

Grammatical Analysis ◽

Machine Translation System ◽

Language Quality ◽

Lexical Errors ◽

Fluent Language

The article presents a rule-based machine translation system from Northern Sami to Norwegian. The grammatical analysis is done with Giellatekno and Divvun's North Sami program for analysis and translation. We have written the transfer component (transfer lexicon and grammatical rules) within the framework of the open machine translation system Apertium. The article contains an evaluation of translated text for two different domains. The translated texts score better on the presentation of the content than on fluent language. By classifying the errors into lexical, grammatical and pragmatic errors, we show that lexical errors are the most harmful for text comprehension. The other two types of errors give a poor language quality, but they have little effect on comprehension. The type of error that is the easiest to correct is the lexical, which is a promising conclusion for the development of a machine translation system for text comprehension.

Download Full-text

What Do Professional Translators Do when Post-editing for the First Time? First Insight into the Spanish-Basque Language Pair

HERMES - Journal of Language and Communication in Business ◽

10.7146/hjlcb.v0i56.97207 ◽

2017 ◽

pp. 89

Author(s):

Nora Aranberri

Keyword(s):

Machine Translation ◽

Training Material ◽

Initial Analysis ◽

Moderate Quality ◽

Editing Activity ◽

Translation Errors ◽

Training And Support ◽

First Time ◽

Insight Into ◽

Language Pair

Machine translation post-editing is becoming commonplace and professional translators are often faced with this unknown task with little training and support. Given the different translation processes involved during post-editing, research suggests that untrained translators do not necessarily make good post-editors. Besides, the post-editing activity will be largely influenced by numerous aspects related to the technology and texts used. Training material, therefore, will need to be tailored to the particular conditions under which post-editing is bound to happen. In this work, we provide a first attempt to uncover what activity professional translators carry out when working from Spanish into Basque. Our initial analysis reveals that when working with moderate machine translation output post-editing shifts from the task of identifying and fixing errors, to that of “patchwork” where post-editors identify the machine translated elements to reuse and connect them using their own contributions. Data also reveal that they primarily focus on correcting machine translation errors but often fail to restrain themselves from editing correct structures. Both findings have clear implications for training and are a step forward in tailoring sessions specifically for language combinations of moderate quality.

Download Full-text

Five Shades of Noise: Analyzing Machine Translation Errors in User-Generated Text

10.18653/v1/w15-4304 ◽

2015 ◽

Cited By ~ 1

Author(s):

Marlies van der Wees ◽

Arianna Bisazza ◽

Christof Monz

Keyword(s):

Machine Translation ◽

Translation Errors

Download Full-text

Correct-and-Memorize: Learning to Translate from Interactive Revisions

Proceedings of the Twenty-Eighth International Joint Conference on Artificial Intelligence ◽

10.24963/ijcai.2019/730 ◽

2019 ◽

Cited By ~ 1

Author(s):

Rongxiang Weng ◽

Hao Zhou ◽

Shujian Huang ◽

Lei Li ◽

Yifan Xia ◽

...

Keyword(s):

Machine Translation ◽

State Of The Art ◽

Translation Process ◽

Neural Machine Translation ◽

Human Interactions ◽

Interactive Environment ◽

Novel Method ◽

Critical Revision ◽

Target Languages ◽

Translation Errors

State-of-the-art machine translation models are still not on a par with human translators. Previous work takes human interactions into the neural machine translation process to obtain improved results in target languages. However, not all model--translation errors are equal -- some are critical while others are minor. In the meanwhile, same translation mistakes occur repeatedly in similar context. To solve both issues, we propose CAMIT, a novel method for translating in an interactive environment. Our proposed method works with critical revision instructions, therefore allows human to correct arbitrary words in model-translated sentences. In addition, CAMIT learns from and softly memorizes revision actions based on the context, alleviating the issue of repeating mistakes. Experiments in both ideal and real interactive translation settings demonstrate that our proposed CAMIT enhances machine translation results significantly while requires fewer revision instructions from human compared to previous methods.

Download Full-text

Minimizing the Bag-of-Ngrams Difference for Non-Autoregressive Neural Machine Translation

Proceedings of the AAAI Conference on Artificial Intelligence ◽

10.1609/aaai.v34i01.5351 ◽

2020 ◽

Vol 34 (01) ◽

pp. 198-205

Author(s):

Chenze Shao ◽

Jinchao Zhang ◽

Yang Feng ◽

Fandong Meng ◽

Jie Zhou

Keyword(s):

Machine Translation ◽

Cross Entropy ◽

Sequential Dependency ◽

Weak Correlation ◽

Neural Machine Translation ◽

Translation Quality ◽

Word Level ◽

Translation Errors ◽

Training Objective ◽

Target Side

Non-Autoregressive Neural Machine Translation (NAT) achieves significant decoding speedup through generating target words independently and simultaneously. However, in the context of non-autoregressive translation, the word-level cross-entropy loss cannot model the target-side sequential dependency properly, leading to its weak correlation with the translation quality. As a result, NAT tends to generate influent translations with over-translation and under-translation errors. In this paper, we propose to train NAT to minimize the Bag-of-Ngrams (BoN) difference between the model output and the reference sentence. The bag-of-ngrams training objective is differentiable and can be efficiently calculated, which encourages NAT to capture the target-side sequential dependency and correlates well with the translation quality. We validate our approach on three translation tasks and show that our approach largely outperforms the NAT baseline by about 5.0 BLEU scores on WMT14 En↔De and about 2.5 BLEU scores on WMT16 En↔Ro.

Download Full-text