Building Machine Translation Systems for the Next Thousand Languages

By Ankur Bapna, Isaac Caswell, Julia Kreutzer, Orhan Firat, Daan van Esch, Aditya Siddhant, Mengmeng Niu, Pallavi N. Baljekar, Xavier García, Wolfgang Macherey and 14 more. 2022

120 citations recorded by the source corpus.

In this paper we share findings from our effort to build practical machine translation (MT) systems capable of translating across over one thousand languages. We describe results in three research domains: (i) Building clean, web-mined datasets for 1500+ languages by leveraging semi-supervised pre-training for language identification and developing data-driven filtering techniques; (ii) Developing practical MT models for under-served languages by leveraging massively multilingual models trained with supervised parallel data for over 100 high-resource languages and monolingual datasets for an additional 1000+ languages; and (iii) Studying the limitations of evaluation metrics for these languages and conducting qualitative analysis of the outputs from our MT models, highlighting several frequent error modes of these types of models. We hope that our work provides useful insights to practitioners working towards building MT systems for currently understudied languages, and highlights research directions that can complement the weaknesses of massively multilingual models in data-sparse settings.

Subfields: Artificial Intelligence, Machine Translation

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

1. Quoc V. Le
2. Graham Neubig
3. Yu Zhang
4. Luke S. Zettlemoyer
5. Mohit Bansal
6. Jinyu Li
7. Jianfeng Gao
8. Kyunghyun Cho

In Machine Translation

1. Ankur Bapna
2. Orhan Firat
3. Yossi Adi
4. Colin Cherry
5. J. Clark
6. Melvin Johnson
7. John Wieting
8. Yu Zhang

Citations in the corpus

• PaLM 2 Technical Report (2023, methodology): “We evaluate translation into English from 26 source languages at different resource levels, including very low resource languages that are underrepresented in digital spaces (Bapna et al., 2022).”
• Low-Resource Languages Jailbreak GPT-4 (2023, methodology): “Google also introduced methods to enable translation for thousands of languages without needing parallel corpora training sets [5].”
• Neural Machine Translation for Low-resource Languages: A Survey (2021, methodology): “[15] used the MASS objective to train a semi-supervised multi-NMT model with massively multilingual parallel data in 112 languages and a 1000-language monolingual corpus.”
• Scaling Speech Technology to 1, 000+ Languages (2023, background): “Multilingual machine translation models have also been scaled to 200 languages [NLLB_Team et al., 2022] and even 1,000 languages [Bapna et al., 2022a].”
• Investigating the Translation Performance of a Large Multilingual Language Model: the Case of BLOOM (2023, unknown): “…introduced in (Dong et al., 2015; Fi-rat et al., 2016; Luong et al., 2016) with limited language coverage, the latest versions of these approaches are able to handle hundreds of languages, including very low-resource language pairs (Fan et al., 2021; Bapna et al., 2022; Costa-juss`a et al., 2022).”
• Google USM: Scaling Automatic Speech Recognition Beyond 100 Languages (2023, methodology): “Web-NTL: For pre-training with unlabeled text, we use a web-crawled corpus of monolingual text containing over 28B sentences [76].”