Exploring Document-Level Literary Machine Translation with Parallel Paragraphs from World Literature

Thai, Katherine; Karpinska, Marzena; Krishna, Kalpesh; Ray, Bill; Inghilleri, Moira; Wieting, John; Iyyer, Mohit

doi:10.18653/v1/2022.emnlp-main.672

Cited by 6 publications

(8 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…At the most basic, and where LLMs seem at present to be most reliable, they can be used to improve the quality of, in ESPL's case, written English. They can also provide translations into English that are better than many non-linguists can achieve even if the resultant text is sub-optimal when evaluated by experts (Thai et al, 2022). Most of us fortunate enough to be Anglophone have never experienced the disadvantage of having to write scientific text that is grammatically and technically correct in a foreign language.…”

Section: Ai Llms and Authorshipmentioning

confidence: 99%

Editorial 2024: Large language models, artificial intelligence and geomorphology

Lane

2024

Earth Surf Processes Landf

View full text Add to dashboard Cite

Large language models (LLMs), such as ChatGPT, have seen an explosion of interest over the last 12 months. It is hard to see an area of education and research where we are not having to face the challenge they pose, a form of artificial intelligence (AI) that is capable of generating text in response to relatively minimum prompting. The text generated can be surprisingly convincing. We now have definitive evidence of work being submitted to Earth Surface Processes and Landforms (ESPL) that has made use of LLMs. As a result, the Editorial Board has recently developed its policy on the use of LLMs in authoring, reviewing and editing papers. In developing our policy, we have taken note that through the journal's publisher Wiley, we are a signatory to the Committee on Publication Ethics (COPE, https://publicationethics.org/). COPE provides us with a framework within which our policy sits, one that given the speed with which LLMs are developing means that it will not be definitive and will surely need to be revisited over the not‐too‐distant future. This extended editorial explains our policy, specifies guidelines for authors, reviewers and editors and also reflects more generally on the positive opportunities and challenges that emerging forms of AI pose to our community and our stated goal of publishing research in geomorphology that is of the highest quality.

show abstract

Section: Ai Llms and Authorshipmentioning

confidence: 99%

Editorial 2024: Large language models, artificial intelligence and geomorphology

Lane

2024

Earth Surf Processes Landf

View full text Add to dashboard Cite

show abstract

“…Although this risk also exists for the human interpreter, in case of uncertainty about the word used, or doubts arising from a lack of connection between the word/phrase and the context, the interpreter can ask follow-up questions, request repetition, rephrasing, additional description. With machine interpreters, there is a risk of misinter-preting the message, partly because the machine translator understands the context in which the translation is taking place and the nuances involved less well than a human interpreter (Thai, 2022). In this case, if the people involved in the activity do not understand both languages, at least at a basic level, they will not be able to check the accuracy and quality of the interpretation and raise objections or clarify doubts.…”

Section: Opportunities and Threatsmentioning

confidence: 99%

Lost in translation: Implementation of the right to a translator through the use of machine translators in the light of EU and Polish Law

Kiejnich-Kruk

2024

10.14746/rpeis

View full text Add to dashboard Cite

Both national and EU law provide for the right of an accused person to receive essential procedural documents in a language they understand and to participate in the proceedings in the presence of an interpreter. The need to implement these procedural guarantees raises several problems, such as the limited availability of interpreters and the cost to the State. This prompts the search for other solutions to implement the indicated individual rights. This article aims to answer the question of the admissibility and legitimacy of using machine translators to implement the right to an interpreter for defendants in criminal proceedings, and to indicate the opportunities and risks associated with the use of AI-based systems in this regard. The article presents the result of a linguistic, systematic and purposive interpretation of the relevant provisions of national laws and the provisions of Directive 2010/64/EU. Formal-dogmatic and statistical methods were used. The results of research into the quality of machine translations and statistics on the costs of translations are presented. The results support the recognition of the admissibility and legitimacy of the use of machine translators in criminal proceedings in order to implement the indicated rights. The text presents possible ways of applying these solutions in judicial practice and formulates a demand to the Ministry of Justice to begin work on the programme without delay. However, currently, both law enforcement and justice authorities can use publicly available machine translation software.

show abstract

“…Jiang et al (2023) curated Bilingual Web Books (BWB), a document-level parallel corpus consisting of 9.6 million sentences and 196 thousand documents (chapters) sourced from English translations of Chinese web novels. Thai et al (2022) introduced PAR3, a multilingual dataset of non-English novels from the public domain, which is aligned at the paragraph level based on both human and automatic translations. Using automatic sentence alignments, Al Ghussin et al ( 2023) extracted parallel paragraphs from Paracrawl (Bañón et al, 2020), which consists of crawled webpages.…”

Section: Datasetsmentioning

confidence: 99%

“…As such, paragraph-level parallel data (Figure 2) may be more suited for document-level NMT and provide richer contextual training signals. Recent work have turned toward literary translation as a challenging, realistic setting for document-level translation Thai et al, 2022;Karpinska and Iyyer, 2023), given that literary texts typically contain complex discourse structures that mandate a document-level frame of reference. As Figure 2 illustrates, sentence alignment is not always feasible when translating literature.…”

Section: Para2para Translationmentioning

confidence: 99%

“…The PARA2PARA dataset consists of 10,545 parallel paragraphs across six novels from the public domain. 9 To our knowledge, the only other paragraph-aligned, parallel dataset sourcing from the literary domain is PAR3 (Thai et al, 2022), which uses Google Translate and fine-tuned GPT-3 (Brown et al, 2020) contrast, the source and target paragraphs in our dataset are culled from professional translations. 10 We then benchmark the dataset under two experimental settings for Zh→En translation: i).…”

Section: Para2para Translationmentioning

confidence: 99%

See 1 more Smart Citation

Challenges in Context-Aware Neural Machine Translation

Jin,

He,

May

et al. 2023

Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing

View full text Add to dashboard Cite

Context-aware neural machine translation, a paradigm that involves leveraging information beyond sentence-level context to resolve intersentential discourse dependencies and improve document-level translation quality, has given rise to a number of recent techniques. However, despite well-reasoned intuitions, most context-aware translation models yield only modest improvements over sentence-level systems. In this work, we investigate and present several core challenges, relating to discourse phenomena, context usage, model architectures, and document-level evaluation, that impede progress within the field. To address these problems, we propose a more realistic setting for document-level translation, called paragraphto-paragraph (PARA2PARA) translation, and collect a new dataset of Chinese-English novels to promote future research. 1

show abstract

Exploring Document-Level Literary Machine Translation with Parallel Paragraphs from World Literature

Cited by 6 publications

References 0 publications

Editorial 2024: Large language models, artificial intelligence and geomorphology

Editorial 2024: Large language models, artificial intelligence and geomorphology

Lost in translation: Implementation of the right to a translator through the use of machine translators in the light of EU and Polish Law

Challenges in Context-Aware Neural Machine Translation

Contact Info

Product

Resources

About