A transfer learning based approach for pronunciation scoring

TL;DR

This paper introduces a transfer learning approach that adapts ASR models for pronunciation scoring, significantly improving accuracy over existing methods by leveraging native speech data and addressing data scarcity issues.

Contribution

The study presents a novel transfer learning method for pronunciation scoring that outperforms state-of-the-art GOP systems, especially in low correction rate scenarios.

Findings

20% improvement over GOP system on EpaDB

Effective adaptation of ASR models for pronunciation scoring

Analysis of design choices impacts

Abstract

Phone-level pronunciation scoring is a challenging task, with performance far from that of human annotators. Standard systems generate a score for each phone in a phrase using models trained for automatic speech recognition (ASR) with native data only. Better performance has been shown when using systems that are trained specifically for the task using non-native data. Yet, such systems face the challenge that datasets labelled for this task are scarce and usually small. In this paper, we present a transfer learning-based approach that leverages a model trained for ASR, adapting it for the task of pronunciation scoring. We analyze the effect of several design choices and compare the performance with a state-of-the-art goodness of pronunciation (GOP) system. Our final system is 20% better than the GOP system on EpaDB, a database for pronunciation scoring research, for a cost function that prioritizes low rates of unnecessary corrections.