Loading [MathJax]/extensions/MathMenu.js
Natural Language Supervision For General-Purpose Audio Representations | IEEE Conference Publication | IEEE Xplore

Natural Language Supervision For General-Purpose Audio Representations


Abstract:

Audio-Language models jointly learn multimodal text and audio representations that enable Zero-Shot inference. Models rely on the encoders to create powerful representati...Show More

Abstract:

Audio-Language models jointly learn multimodal text and audio representations that enable Zero-Shot inference. Models rely on the encoders to create powerful representations of the input and generalize to multiple tasks ranging from sounds, music, and speech. Although models have achieved remarkable performance, there is still a gap with task-specific models. In this paper, we propose a Contrastive Language-Audio Pretraining model that is pretrained with a diverse collection of 4.6M audio-text pairs employing two innovative encoders for Zero-Shot inference. To learn audio representations, we trained an audio encoder on 22 audio tasks, instead of the standard training of sound event classification. To learn language representations, we trained an autoregressive decoder-only model instead of the standard encoder-only models. Then, the audio and language representations are brought into a joint multimodal space using Contrastive Learning. We used our encoders to improve the downstream performance by a large margin. We extensively evaluated the generalization of our representations on 26 downstream tasks, the largest in the literature. Our model achieves state of the art results in several tasks outperforming 4 different models and leading the way towards general-purpose audio representations. Code is on GitHub1.
Date of Conference: 14-19 April 2024
Date Added to IEEE Xplore: 18 March 2024
ISBN Information:

ISSN Information:

Conference Location: Seoul, Korea, Republic of

Contact IEEE to Subscribe

References

References is not available for this document.