Reformulating Speaker Diarization as Community Detection With Emphasis On Topological Structure

TL;DR

This paper redefines speaker diarization as a community detection problem, leveraging topological structure and novel techniques to significantly improve clustering accuracy and reduce diarization error rate.

Contribution

It introduces a community detection framework for speaker diarization, outperforming traditional clustering methods and integrating topological analysis with an end-to-end system.

Findings

Leiden algorithm outperforms previous clustering methods

Dimensionality reduction preserves topological structure

End-to-end system achieves up to 70% DER reduction

Abstract

Clustering-based speaker diarization has stood firm as one of the major approaches in reality, despite recent development in end-to-end diarization. However, clustering methods have not been explored extensively for speaker diarization. Commonly-used methods such as k-means, spectral clustering, and agglomerative hierarchical clustering only take into account properties such as proximity and relative densities. In this paper we propose to view clustering-based diarization as a community detection problem. By doing so the topological structure is considered. This work has four major contributions. First it is shown that Leiden community detection algorithm significantly outperforms the previous methods on the clustering of speaker-segments. Second, we propose to use uniform manifold approximation to reduce dimension while retaining global and local topological structure. Third, a masked filtering approach is introduced to extract "clean" speaker embeddings. Finally, the community structure is applied to an end-to-end post-processing network to obtain diarization results. The final system presents a relative DER reduction of up to 70 percent. The breakdown contribution of each component is analyzed.