State of the art end-to-end automatic speech recognition (ASR) models require large speech datasets for training. The Mozilla Common Voice project crowd-sources read speech to address this need. However, this approach often results in many audio utterances being recorded for each written sentence. Using Kiswahili speech data, this paper first explores how much audio repetition in utterances is permissible in a training set before model degradation occurs, then examines the extent to which audio augmentation techniques can be employed to increase the diversity of speech characteristics and improve accuracy. We find that repetition up to a ratio of 1 sentence to 8 audio recordings improves performance, but performance degrades at a ratio of 1:16. We also find small improvements from frequency mask, time mask and tempo augmentation. Our findings provide guidance on training set construction for ASR practitioners, particularly those working in under-served languages.

Language technologies, particularly speech technologies, are becoming more pervasive for access to digital platforms and resources. This brings to the forefront concerns of their inclusivity, first in terms of language diversity. Additionally, research shows speech recognition to be more accurate for men than for women and more accurate for individuals younger than 30 years of age than those older. In the Global South where languages are low resource, these same issues should be taken into consideration in data collection efforts to not replicate these mistakes. It is also important to note that in varying contexts within the Global South, this work presents additional nuance and potential for bias based on accents, related dialects and variants of a language. This paper documents i) the designing and execution of a Linguists Engagement for purposes of building an inclusive Kiswahili Speech Recognition dataset, representative of the diversity among speakers of the language ii) the unexpected yet key learning in terms of socio-linguistcs which demonstrate the importance of multi-disciplinarity in teams developing datasets and NLP technologies iii) the creation of a test dataset intended to be used for evaluating the performance of Speech Recognition models on demographic groups that are likely to be underrepresented.

Research in NLP lacks geographic diversity, and the question of how NLP can be scaled to low-resourced languages has not yet been adequately solved. ‘Low-resourced’-ness is a complex problem going beyond data availability and reflects systemic problems in society. In this paper, we focus on the task of Machine Translation (MT), that plays a crucial role for information accessibility and communication worldwide. Despite immense improvements in MT over the past decade, MT is centered around a few high-resourced languages. As MT researchers cannot solve the problem of low-resourcedness alone, we propose participatory research as a means to involve all necessary agents required in the MT development process. We demonstrate the feasibility and scalability of participatory research with a case study on MT for African languages. Its implementation leads to a collection of novel translation datasets, MT benchmarks for over 30 languages, with human evaluations for a third of them, and enables participants without formal training to make a unique scientific contribution. Benchmarks, models, data, code, and evaluation results are released at https://github.com/masakhane-io/masakhane-mt.

As language and speech technologies become more advanced, the lack of fundamental digital resources for African languages, such as data, spell checkers and PoS taggers, means that the digital divide between these languages and others keeps growing. This work details the organisation of the AI4D - African Language Dataset Challenge, an effort to incentivize the creation, curation and uncovering to African language datasets through a competitive challenge, particularly datasets that are annotated or prepared for use in a downstream NLP task.