Learning Contextual Tag Embeddings for Cross-Modal Alignment of Audio and Tags

Self-supervised audio representation learning offers an attractive alternative for obtaining generic audio embeddings, capable to be employed into various downstream tasks. Published approaches that consider both audio and words/tags associated with audio do not employ text processing models that are capable to generalize to tags unknown during training. In this work we propose a method for learning audio representations using an audio autoencoder (AAE), a general word embed-dings model (WEM), and a multi-head self-attention (MHA) mechanism. MHA attends on the output of the WEM, pro-viding a contextualized representation of the tags associated with the audio, and we align the output of MHA with the out-put of the encoder of AAE using a contrastive loss. We jointly optimize AAE and MHA and we evaluate the audio representations (i.e. the output of the encoder of AAE) by utilizing them in three different downstream tasks, namely sound, music genre, and music instrument classification. Our results show that employing multi-head self-attention with multiple heads in the tag-based network can induce better learned audio representations.

1. INTRODUCTION

An effective way to learn audio representations that can be used for sound classification involves training deep neural networks (DNNs) on supervised tasks, using large annotated datasets [1], [2], [3]. However, these datasets require a considerable amount of effort to be built and are always limited in size, hindering the performance of learned representations. Recent research approaches explore and adopt unsupervised, self-supervised or semi-supervised learning methods for obtaining generic audio representations, that later can be used for different downstream tasks [4], [5], [6], [7]. The large amount of multimedia data available online is a great opportunity for these types of approaches to learn powerful audio representations.