Audio-Language Models for Audio-Centric Tasks: A Systematic Survey

TL;DR

This systematic survey reviews Audio-Language Models (ALMs), highlighting their architectures, training methods, and applications across audio domains, emphasizing their zero-shot capabilities and identifying future research directions.

Contribution

The paper provides the first comprehensive survey of ALMs, including a unified taxonomy and analysis of their development, limitations, and future trends in audio-centric tasks.

Findings

ALMs demonstrate strong zero-shot and generalization abilities.

A unified taxonomy of ALM architectures and training objectives.

Identification of current limitations and promising future research directions.

Abstract

Audio-Language Models (ALMs), trained on paired audio-text data, are designed to process, understand, and reason about audio-centric multimodal content. Unlike traditional supervised approaches that use predefined labels, ALMs leverage natural language supervision to better handle complex real-world audio scenes with multiple overlapping events. While demonstrating impressive zero-shot and task generalization capabilities, there is still a notable lack of systematic surveys that comprehensively organize and analyze developments. In this paper, we present the first systematic review of ALMs with three main contributions: (1) comprehensive coverage of ALM works across speech, music, and sound from a general audio perspective; (2) a unified taxonomy of ALM foundations, including model architectures and training objectives; (3) establishment of a research landscape capturing mutual promotion and constraints among different research aspects, aiding in summarizing evaluations, limitations, concerns and promising directions. Our review contributes to helping researchers understand the development of existing technologies and future trends, while also providing valuable references for implementation in practical applications.