On the use of Performer and Agent Attention for Spoken Language Identification

TL;DR

This paper investigates the effectiveness of Performer and Agent-Attention mechanisms combined with statistical pooling for spoken language identification, demonstrating that Performer-Attention outperforms traditional self-attention with less computational cost.

Contribution

It introduces and evaluates Performer and Agent-Attention mechanisms in LID, showing their advantages over standard self-attention in accuracy and efficiency.

Findings

Performer-Attention outperforms self-attention in LID tasks.

Agent-Attention performs comparably or better than self-attention.

Performer-Attention is more computationally efficient.

Abstract

One of the methods for language Identification (LID) involves deriving speech representation from pre-trained models using self-supervised learning, followed by fine-tuning the model for the LID task. State-of-the-art approaches for LID use an attention-based statistical pooling layer to facilitate the aggregation of contextual information across time frames of the embedding vectors extracted from the pre-trained model. In this paper, we delve into exploring recently proposed attention mechanisms, namely performer and agent-attention, in conjunction with the statistical pooling layer. The LID experiments are performed on three datasets: VoxPopuli, FLEURS, and VoxLingua. We compare their performance against vanilla self-attention. Our findings suggest that performer-attention outperforms self-attention and agent-attention exhibits comparable or occasionally superior performance to self-attention, while also being computationally less expensive.