TW-Sound580K: A Regional Audio-Text Dataset with Verification-Guided Curation for Localized Audio-Language Modeling

TL;DR

This paper introduces TW-Sound580K, a large Taiwanese audio-text dataset created with a verification-guided pipeline, and demonstrates its effectiveness in improving localized audio-language modeling with a new Tai-LALM system.

Contribution

The paper presents a novel dataset curation method using Verify-Generate-Critique and Dual-ASR validation, along with a specialized LALM model for regional dialects.

Findings

Tai-LALM achieves 49.1% accuracy on TAU Benchmark.

The dataset improves LALM performance over zero-shot baselines.

Dynamic Dual-ASR arbitration enhances transcription quality.

Abstract

Large Audio-Language Models (LALMs) typically struggle with localized dialectal prosody due to the scarcity of specialized corpora. We present TW-Sound580K, a Taiwanese audio-text instruction dataset developed through a Verify-Generate-Critique (VGC) protocol. This pipeline leverages Dual-ASR validation to filter 522K raw clips, subsequently expanding them into 580,000 high-fidelity instruction pairs using a teacher model. The dataset's utility is demonstrated through Tai-LALM, which fine-tunes a DeSTA 2.5-Audio-initialized backbone and incorporates a dynamic Dual-ASR Arbitration strategy to optimize transcription selection during inference. On the TAU Benchmark, Tai-LALM reaches 49.1% accuracy, marking a 6.5% absolute improvement over the zero-shot baseline (42.6% with ASR text conditioning). This confirms that integrating regional corpora with rigorous curation and dynamic arbitration significantly enhances LALM performance on localized speech.