Lexicon Stratification for Translating Out-of-Vocabulary Words

Tsvetkov, Yulia; Dyer, Chris

doi:10.3115/v1/p15-2021

Cited by 15 publications

(13 citation statements)

References 19 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Finally, as observed in the previous section, there are many Out of Vocabulary OOV words in any language in the world and many researchers, including Tsvetkov & Dyer (2015), concede that transliteration should be the way to go when attempting to translate texts that contain these OOVs. The same can be applied to proper nouns according to Habash (2008) who also believes that transliteration is the proper course of action when it comes to translating texts containing such words.…”

Section: Transliteration and Arabizationmentioning

confidence: 81%

“…According to Tsvetkov & Dyer (2015), transliterated words are among the four categories of vocabulary in a language and they call these words unassimilated. The remaining three being core words of the language, assimilated and semi-assimilated.…”

Section: Transliteration and Arabizationmentioning

confidence: 99%

“…I observed that the majority of recent research in the fields of OOV and transliteration (e.g; Chalabi, Morsy, Awadalla, El-Sharqwi, & Hassan, 2015;Habash, 2008;Tsvetkov & Dyer, 2015, to name a few) is more concerned with machine translation not the actual process of transliteration of OOV words. I struggled to find any recent study that investigates the impact of transliteration on the final product and readers" perceptions of the translated/transliterated text, which is important given the fact that the purpose of translating a text, automatically or otherwise, is to produce relevant texts in the potential readers" language that is easier to understand compared to the original text in the source language, usually English.…”

Section: Transliteration and Arabizationmentioning

confidence: 99%

See 2 more Smart Citations

Translation vs. Transliteration: Arabization in Scientific Texts

Grami

2019

JELTL

View full text Add to dashboard Cite

This paper looks at the concepts of translation and transliteration in general and in scientific and academic texts in particular. In simple terms, the former refers to the process of finding equivalents in the target language (as opposed to the original language of the text), while the latter refers to writing the original word using the characters of the target language. The paper argues that translation works well in texts that explain, describe, detail, instruct and summarize while transliteration works better in concepts, processes, known procedures and proper nouns, to mention but a few. The paper suggests that the reliance on literal translation of terms and concepts can be counterproductive to the purpose of translation. Six computer science students were involved in a small-scale experiment. Tests were designed to determine which approach, Arabization or literal translation, is more efficient by measuring the time students took to complete certain tasks and whether students can trace the translated word back to its English origin. All participants were interviewed afterwards. Results showed that they preferred transliterated terms and that Arabic literal translation was not helpful. Results also showed that transliteration of scientific texts helped students understand faster and more accurately. The paper recommends a hybrid approach that employs both methods depending on what terms or processes are being translated.

show abstract

Section: Transliteration and Arabizationmentioning

confidence: 81%

Section: Transliteration and Arabizationmentioning

confidence: 99%

Section: Transliteration and Arabizationmentioning

confidence: 99%

See 1 more Smart Citation

Translation vs. Transliteration: Arabization in Scientific Texts

Grami

2019

JELTL

View full text Add to dashboard Cite

show abstract

“…Lexical borrowing has received relatively little attention in natural language processing area. Tsvetkov and Dyer [7] proposed a morph-phonological transformation model to obtain good performance at predicting donor forms from borrowed forms. Tsvetkov et al [7] suggested to use the lexical borrowing as a model in an SMT framework to translate OOV words.…”

Section: Loanword Identificationmentioning

confidence: 99%

Improving Loanword Identification in Low-Resource Language with Data Augmentation and Multiple Feature Fusion

Zhu

Nie

2021

Computational Intelligence and Neuroscience

View full text Add to dashboard Cite

Loanword identification is studied in recent years to alleviate data sparseness in several natural language processing (NLP) tasks, such as machine translation, cross-lingual information retrieval, and so on. However, recent studies on this topic usually put efforts on high-resource languages (such as Chinese, English, and Russian); for low-resource languages, such as Uyghur and Mongolian, due to the limitation of resources and lack of annotated data, loanword identification on these languages tends to have lower performance. To overcome this problem, we first propose a lexical constraint-based data augmentation method to generate training data for low-resource language loanword identification; then, a loanword identification model based on a log-linear RNN is introduced to improve the performance of low-resource loanword identification by incorporating features such as word-level embeddings, character-level embeddings, pronunciation similarity, and part-of-speech (POS) into one model. Experimental results on loanword identification in Uyghur (in this study, we mainly focus on Arabic, Chinese, Russian, and Turkish loanwords in Uyghur) showed that our proposed method achieves best performance compared with several strong baseline systems.

show abstract

“…Cognates and the problem of cognate identification have been extensively studied in the fields of language typology and historical linguistics, as cognates are considered useful for researching the relatedness of languages (Bhattacharya et al, 2018). Cognates are also used in computational linguistics, e.g., for lexicon extension (Wu and Yarowsky, 2018) or to improve cross-lingual NLP tasks such as machine translation or bilingual word recognition (Kondrak et al, 2003;Tsvetkov and Dyer, 2015).…”

Section: Introductionmentioning

confidence: 99%

CogNet: A Large-Scale Cognate Database

Batsuren¹,

Bella²,

Giunchiglia³

2019

Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics

View full text Add to dashboard Cite

This paper introduces CogNet, a new, large-scale lexical database that provides cognates-words of common origin and meaning-across languages. The database currently contains 3.1 million cognate pairs across 338 languages using 35 writing systems. The paper also describes the automated method by which cognates were computed from publicly available wordnets, with an accuracy evaluated to 94%. Finally, statistics and early insights about the cognate data are presented, hinting at a possible future exploitation of the resource 1 by various fields of lingustics.

show abstract

Lexicon Stratification for Translating Out-of-Vocabulary Words

Cited by 15 publications

References 19 publications

Translation vs. Transliteration: Arabization in Scientific Texts

Translation vs. Transliteration: Arabization in Scientific Texts

Improving Loanword Identification in Low-Resource Language with Data Augmentation and Multiple Feature Fusion

CogNet: A Large-Scale Cognate Database

Contact Info

Product

Resources

About