Paraphrasing Training Data for Statistical Machine Translation

Nichols, Eric; Bond, Francis; Appling, Darren Scott; Matsumoto, Yūji

doi:10.5715/jnlp.17.3_101

Cited by 5 publications

(4 citation statements)

References 11 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This method was done by just using the paraphrases as a means for data augmentation on the source side, such as reported by Nichols et al (2010), to leverage SMT systems [19]. All of the paraphrases and their original sentence were combined, and the target sentence was duplicated by the number of multiple paraphrases.…”

Section: Combining All Data In a Single Modelmentioning

confidence: 99%

Leveraging Neural Caption Translation with Visually Grounded Paraphrase Augmentation

Effendi

Sakti

Sudoh

et al. 2020

IEICE Trans. Inf. & Syst.

View full text Add to dashboard Cite

Since a concept can be represented by different vocabularies, styles, and levels of detail, a translation task resembles a many-to-many mapping task from a distribution of sentences in the source language into a distribution of sentences in the target language. This viewpoint, however, is not fully implemented in current neural machine translation (NMT), which is one-to-one sentence mapping. In this study, we represent the distribution itself as multiple paraphrase sentences, which will enrich the model context understanding and trigger it to produce numerous hypotheses. We use a visually grounded paraphrase (VGP), which uses images as a constraint of the concept in paraphrasing, to guarantee that the created paraphrases are within the intended distribution. In this way, our method can also be considered as incorporating image information into NMT without using the image itself. We implement this idea by crowdsourcing a paraphrasing corpus that realizes VGP and construct neural paraphrasing that behaves as expert models in a NMT. Our experimental results reveal that our proposed VGP augmentation strategies showed improvement against a vanilla NMT baseline.

show abstract

Section: Combining All Data In a Single Modelmentioning

confidence: 99%

Leveraging Neural Caption Translation with Visually Grounded Paraphrase Augmentation

Effendi

Sakti

Sudoh

et al. 2020

IEICE Trans. Inf. & Syst.

View full text Add to dashboard Cite

show abstract

“…Relevant studies have also confirmed that the use of CL rules can have a positive impact on MT output (Aikawa et al 2007; Bernth 1998; Mitamura and Nyberg 1995; Mitamura 1999). One of the objectives of this study is to summarize a set of language control rules, which can be used to retell or edit the source language sentences before translation, reduce ambiguity and improve the quality of machine translation [3,4,5]. The steps of designing controlled language rules are as follows:…”

Section: Controlling Languagementioning

confidence: 99%

Research on High Quality Bilingual Parallel Documents of Science and Technology Produced by Application Software Tools and Control Language

Jing¹,

Zhu²

2019

dtem

View full text Add to dashboard Cite

The aim of this study is to help English-based technicians produce high-quality translated documents with the help of free software tools (such as machine translation+post-translation compilation technology, computer-aided translation technology, etc.). This paper discusses a method of iteratively rewriting technical documents to help technicians generate more accurate bilingual parallel documents. It also studies how to use control language rules to rewrite technical documents to eliminate ambiguity and sentence complexity in order to improve the translatability of documents. Finally, a method is proposed to help users choose the most suitable machine translation engine.

show abstract

“…As shown in Fig 2, the multiple pivot method is used to obtain the candidate paraphrase process, which is assumed to be composed of N translation engines, and assumes that each engine can handle M languages. The system can input any source language fragment S through any of the N inputs, and then obtain the representation of the target language fragment T from any of the N output engines [9,10]. The multiple pivot based paraphrase generation method has the following two advantages.…”

Section: Retrieval Of Rehearsal Resources Based On Multimentioning

confidence: 99%

The Application of Paraphrasing Technology of Machine Translation in the Construction of Corpus

Jing¹

2017

Proceedings of the 2017 2nd International Conference on Education, Management Science and Economics (ICEMSE 2017)

View full text Add to dashboard Cite

At present, with the updating of knowledge and information, bilingual libraries are unlikely to contain all linguistic phenomena. In the course of Machine Translation, it is impossible to process unknown texts that are not present in the corpus. With paraphrasing technology, the unknown text fragments can be translated into idioms with relatively close meaning. The goal of this paper is to combine multiple resources to improve the accuracy and comprehensive performance of paraphrasing technology. This can solve the problem of data sparseness and make the expression more diversified. When bilingual corpus is enriched, the expressions are diversified; the accuracy of machine translation is improved.

show abstract

Paraphrasing Training Data for Statistical Machine Translation

Cited by 5 publications

References 11 publications

Leveraging Neural Caption Translation with Visually Grounded Paraphrase Augmentation

Leveraging Neural Caption Translation with Visually Grounded Paraphrase Augmentation

Research on High Quality Bilingual Parallel Documents of Science and Technology Produced by Application Software Tools and Control Language

The Application of Paraphrasing Technology of Machine Translation in the Construction of Corpus

Contact Info

Product

Resources

About