Daan van Esch

TL;DR: Results in three research domains are described, which include building clean, web-mined datasets for 1500+ languages by leveraging semi-supervised pre-training for language identification and developing data-drivenData-driven language identification techniques and developing practical MT models for under-served languages.

TL;DR: Two classes of techniques are proposed: wordlist-based tunable-precision filters and transformer-based semi-supervised LangID models, which increase median dataset precision from 5.5% to 71.2% and enable an initial data set covering 100K or more relatively clean sentences in each of 500+ languages, paving the way towards a 1,000-language web text corpus.

TL;DR: The development of Elpis is described, a pipeline which language documentation workers with minimal computational experience can use to build their own speech recognition models, resulting in models being built for 16 languages from the Asia-Pacific region.

TL;DR: In this paper, the authors manually audit the quality of 205 language-specific corpora released with five major public datasets (CCAligned, ParaCrawl, WikiMatrix, OSCAR, mC4) and audit the correctness of language codes in a sixth (JW300).

TL;DR: The automated multi-language text normalization infrastructure that prepares textual data to train language models used in Google’s keyboards and speech recognition systems is described, across hundreds of language varieties.