Learning from Limited Labels: Transductive Graph Label Propagation for Indian Music Analysis

TL;DR

This work tackles the lack of large labeled audio datasets by applying transductive label propagation on a graph of audio embeddings to extend sparse labels to vast unlabeled corpora. Focusing on Indian Art Music, the authors validate the approach on Raga identification and Instrument recognition by integrating public datasets with Prasar Bharati Archives, and demonstrate that LP can significantly reduce labeling effort while yielding high-quality pseudo-labels. The results show substantial improvements over fully supervised baselines, with notable gains in file-level accuracy for Raga classification and high accuracy in instrument recognition, highlighting the practical potential of graph-based semi-supervised labeling for scalable MIR in data-scarce domains. This approach promises to democratize data annotation and accelerate progress in music information retrieval, especially for specialized musical traditions.

Abstract

Supervised machine learning frameworks rely on extensive labeled datasets for robust performance on real-world tasks. However, there is a lack of large annotated datasets in audio and music domains, as annotating such recordings is resource-intensive, laborious, and often require expert domain knowledge. In this work, we explore the use of label propagation (LP), a graph-based semi-supervised learning technique, for automatically labeling the unlabeled set in an unsupervised manner. By constructing a similarity graph over audio embeddings, we propagate limited label information from a small annotated subset to a larger unlabeled corpus in a transductive, semi-supervised setting. We apply this method to two tasks in Indian Art Music (IAM): Raga identification and Instrument classification. For both these tasks, we integrate multiple public datasets along with additional recordings we acquire from Prasar Bharati Archives to perform LP. Our experiments demonstrate that LP significantly reduces labeling overhead and produces higher-quality annotations compared to conventional baseline methods, including those based on pretrained inductive models. These results highlight the potential of graph-based semi-supervised learning to democratize data annotation and accelerate progress in music information retrieval.

Figures (3)

• Figure 1: Source-wise duration of all instrument samples used from various datasets (in seconds). PB represents Prasar Bharati audios.
• Figure 2: Flowchart illustrating the Label Propagation process for automatic annotation of unlabeled audio samples. A small labeled dataset is combined with a larger set of partially labeled or unlabeled samples. Label Propagation is then applied using a similarity graph, enabling the transfer of label information from the labeled subset to the unlabeled data, resulting in the entire corpus being annotated in a transductive, semi-supervised manner.
• Figure 3: Model architecture used for the Raga Identification task. The model consists of convolutional layers followed by LSTM layers to capture temporal dependencies in the audio data.