Towards few-shot isolated word reading assessment
This addresses reading assessment for children in low-resource language settings, but the results are incremental as they reveal limitations rather than breakthroughs.
The researchers tackled the problem of isolated word reading assessment in low-resource settings by developing a few-shot ASR-free method that compares child speech to adult reference templates using SSL model encodings. Their experiments on an Afrikaans benchmark showed reasonable performance for adults but a substantial drop for child speech, highlighting SSL limitations in this context.
We explore an ASR-free method for isolated word reading assessment in low-resource settings. Our few-shot approach compares input child speech to a small set of adult-provided reference templates. Inputs and templates are encoded using intermediate layers from large self-supervised learned (SSL) models. Using an Afrikaans child speech benchmark, we investigate design options such as discretising SSL features and barycentre averaging of the templates. Idealised experiments show reasonable performance for adults, but a substantial drop for child speech input, even with child templates. Despite the success of employing SSL representations in low-resource speech tasks, our work highlights the limitations of SSL representations for processing child data when used in a few-shot classification system.