Acoustic Model Optimization over Multiple Data Sources: Merging and Valuation

TL;DR

This paper introduces a novel multi-source acoustic model training paradigm for ASR, combining models trained on different datasets using two algorithms, GMA and SOMA, and evaluates data contribution with Shapley Values.

Contribution

It proposes two new algorithms, GMA and SOMA, for merging acoustic models trained on separate data sources, and applies Shapley Values for data contribution assessment.

Findings

GMA improves model merging quality but is slow.

SOMA maintains accuracy with higher efficiency.

Proposed methods outperform state-of-the-art on public data.

Abstract

Due to the rising awareness of privacy protection and the voluminous scale of speech data, it is becoming infeasible for Automatic Speech Recognition (ASR) system developers to train the acoustic model with complete data as before. For example, the data may be owned by different curators, and it is not allowed to share with others. In this paper, we propose a novel paradigm to solve salient problems plaguing the ASR field. In the first stage, multiple acoustic models are trained based upon different subsets of the complete speech data, while in the second phase, two novel algorithms are utilized to generate a high-quality acoustic model based upon those trained on data subsets. We first propose the Genetic Merge Algorithm (GMA), which is a highly specialized algorithm for optimizing acoustic models but suffers from low efficiency. We further propose the SGD-Based Optimizational Merge Algorithm (SOMA), which effectively alleviates the efficiency bottleneck of GMA and maintains superior model accuracy. Extensive experiments on public data show that the proposed methods can significantly outperform the state-of-the-art. Furthermore, we introduce Shapley Value to estimate the contribution score of the trained models, which is useful for evaluating the effectiveness of the data and providing fair incentives to their curators.