While massively multilingual speech models like wav2vec 2.0 XLSR-128 can be directly fine-tuned for automatic speech recognition (ASR), downstream performance can still be relatively poor on languages that are under-represented in the pre-training data. Continued pre-training on 70–200 hours of untranscribed speech in these languages can help — but what about languages without that much recorded data? For such cases, we show that supplementing the target language with data from a similar, higher-resource ‘donor’ language can help. For example, continued pretraining on only 10 hours of low-resource Punjabi supplemented with 60 hours of donor Hindi is almost as good as continued pretraining on 70 hours of Punjabi. By contrast, sourcing supplemental data from less similar donors like Bengali does not improve ASR performance. To inform donor language selection, we propose a novel similarity metric based on the sequence distribution of induced acoustic units: the Acoustic Token Distribution Similarity (ATDS). Across a set of typologically different target languages (Punjabi, Galician, Iban, Setswana), we show that the ATDS between the target language and its candidate donors precisely predicts target language ASR performance.

Les systèmes de reconnaissance automatique de la parole atteignent désormais des degrés de précision élevés sur la base d’un corpus d’entraînement limité à deux ou trois heures d’enregistrements transcrits (pour un système mono-locuteur). Au-delà de l’intérêt pratique que présentent ces avancées technologiques pour les tâches de documentation de langues rares et en danger, se pose la question de leur apport pour la réflexion du phonéticien/phonologue. En effet, le modèle acoustique prend en entrée des transcriptions qui reposent sur un ensemble d’hypothèses plus ou moins explicites. Le modèle acoustique, décalqué (par des méthodes statistiques) de l’écrit du linguiste, peut-il être interrogé par ce dernier, en un jeu de miroir ? Notre étude s’appuie sur des exemples d’une langue « rare » de la famille sino-tibétaine, le na (mosuo), pour illustrer la façon dont l’analyse d’erreurs permet une confrontation renouvelée avec le signal acoustique.

We present findings from the creation of a massively parallel corpus in over 1600 languages, the Johns Hopkins University Bible Corpus (JHUBC). The corpus consists of over 4000 unique translations of the Christian Bible and counting. Our data is derived from scraping several online resources and merging them with existing corpora, combining them under a common scheme that is verse-parallel across all translations. We detail our effort to scrape, clean, align, and utilize this ripe multilingual dataset. The corpus captures the great typological variety of the world’s languages. We catalog this by showing highly similar proportions of representation of Ethnologue’s typological features in our corpus. We also give an example application: projecting pronoun features like clusivity across alignments to richly annotate languages which do not mark the distinction.

We investigate the problem of searching for a lexeme-set in speech by searching for its inflectional variants. Experimental results indicate how lexeme-set search performance changes with the number of hypothesized inflections, while ablation experiments highlight the relative importance of different components in the lexeme-set search pipeline and the value of using curated inflectional paradigms. We provide a recipe and evaluation set for the community to use as an extrinsic measure of the performance of inflection generation approaches.

We report on adaptation of multilingual end-to-end speech recognition models trained on as many as 100 languages. Our findings shed light on the relative importance of similarity between the target and pretraining languages along the dimensions of phonetics, phonology, language family, geographical location, and orthography. In this context, experiments demonstrate the effectiveness of two additional pretraining objectives in encouraging language-independent encoder representations: a context-independent phoneme objective paired with a language-adversarial classification objective.

Most languages have no established writing system and minimal written records. However, textual data is essential for natural language processing, and particularly important for training language models to support speech recognition. Even in cases where text data is missing, there are some languages for which bilingual lexicons are available, since creating lexicons is a fundamental task of documentary linguistics. We investigate the use of such lexicons to improve language models when textual training data is limited to as few as a thousand sentences. The method involves learning cross-lingual word embeddings as a preliminary step in training monolingual language models. Results across a number of languages show that language models are improved by this pre-training. Application to Yongning Na, a threatened language, highlights challenges in deploying the approach in real low-resource environments.

We address the problem of automatic short answer grading, evaluating a collection of approaches inspired by recent advances in distributional text representations. In addition, we propose an unsupervised approach for determining text similarity using one-to-many alignment of word vectors. We evaluate the proposed technique across two datasets from different domains, namely, computer science and English reading comprehension, that additionally vary between highschool level and undergraduate students. Experiments demonstrate that the proposed technique often outperforms other compositional distributional semantics approaches as well as vector space methods such as latent semantic analysis. When combined with a scoring scheme, the proposed technique provides a powerful tool for tackling the complex problem of short answer grading. We also discuss a number of other key points worthy of consideration in preparing viable, easy-to-deploy automatic short-answer grading systems for the real-world.