With the success of large-scale pre-training and multilingual modeling in Natural Language Processing (NLP), recent years have seen a proliferation of large, Web-mined text datasets covering hundreds of languages. We manually audit the quality of 205 language-specific corpora released with five major public datasets (CCAligned, ParaCrawl, WikiMatrix, OSCAR, mC4). Lower-resource corpora have systematic issues: At least 15 corpora have no usable text, and a significant fraction contains less than 50% sentences of acceptable quality. In addition, many are mislabeled or use nonstandard/ambiguous language codes. We demonstrate that these issues are easy to detect even for non-proficient speakers, and supplement the human audit with automatic analyses. Finally, we recommend techniques to evaluate and improve multilingual corpora and discuss potential risks that come with low-quality data releases.

We describe an open-source dataset providing metadata for about 2,800 language varieties used in the world today. Specifically, the dataset provides the attested writing system(s) for each of these 2,800+ varieties, as well as an estimated speaker count for each variety. This dataset was developed through internal research and has been used for analyses around language technologies. This is the largest publicly-available, machine-readable resource with writing system and speaker information for the world’s languages. We analyze the distribution of languages and writing systems in our data and compare it to their representation in current NLP. We hope the availability of this data will catalyze research in under-represented languages.

The applications of automatic speech recognition (ASR) systems are proliferating, in part due to recent significant quality improvements. However, as recent work indicates, even state-of-the-art speech recognition systems – some which deliver impressive benchmark results, struggle to generalize across use cases. We review relevant work, and, hoping to inform future benchmark development, outline a taxonomy of speech recognition use cases, proposed for the next generation of ASR benchmarks. We also survey work on metrics, in addition to the de facto standard Word Error Rate (WER) metric, and we introduce a versatile framework designed to describe interactions between linguistic variation and ASR performance metrics.

This paper presents a methodology for rapidly generating FST-based verbalizers for ASR and TTS systems by efficiently sourcing language-specific data. We describe a questionnaire which collects the necessary data to bootstrap the number grammar induction system and parameterize the verbalizer templates described in Ritchie et al. (2019), and a machine-readable data store which allows the data collected through the questionnaire to be supplemented by additional data from other sources. This system allows us to rapidly scale technologies such as ASR and TTS to more languages, including low-resource languages.

Large text corpora are increasingly important for a wide variety of Natural Language Processing (NLP) tasks, and automatic language identification (LangID) is a core technology needed to collect such datasets in a multilingual context. LangID is largely treated as solved in the literature, with models reported that achieve over 90% average F1 on as many as 1,366 languages. We train LangID models on up to 1,629 languages with comparable quality on held-out test sets, but find that human-judged LangID accuracy for web-crawl text corpora created using these models is only around 5% for many lower-resource languages, suggesting a need for more robust evaluation. Further analysis revealed a variety of error modes, arising from domain mismatch, class imbalance, language similarity, and insufficiently expressive models. We propose two classes of techniques to mitigate these errors: wordlist-based tunable-precision filters (for which we release curated lists in about 500 languages) and transformer-based semi-supervised LangID models, which increase median dataset precision from 5.5% to 71.2%. These techniques enable us to create 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.