Probing Statistical Representations For End-To-End ASR

TL;DR

This paper analyzes internal neural representations in end-to-end speech recognition models, revealing how layer dependencies influence recognition performance and offering insights for improving model design.

Contribution

It introduces a novel analysis of transformer-based ASR models using SVCCA to understand internal dependencies and their impact on recognition accuracy.

Findings

Neural representations within transformer layers show correlated behavior.

Layer dependencies significantly affect recognition performance.

Insights can guide the development of better end-to-end ASR models.

Abstract

End-to-End automatic speech recognition (ASR) models aim to learn a generalised speech representation to perform recognition. In this domain there is little research to analyse internal representation dependencies and their relationship to modelling approaches. This paper investigates cross-domain language model dependencies within transformer architectures using SVCCA and uses these insights to exploit modelling approaches. It was found that specific neural representations within the transformer layers exhibit correlated behaviour which impacts recognition performance. Altogether, this work provides analysis of the modelling approaches affecting contextual dependencies and ASR performance, and can be used to create or adapt better performing End-to-End ASR models and also for downstream tasks.