Saved in:
Bibliographic Details
Main Authors: Chen, Yu-Wen, Ho, William, Topaz, Maxim, Hirschberg, Julia, Kostic, Zoran
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2509.15082
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911161889325056
author Chen, Yu-Wen
Ho, William
Topaz, Maxim
Hirschberg, Julia
Kostic, Zoran
author_facet Chen, Yu-Wen
Ho, William
Topaz, Maxim
Hirschberg, Julia
Kostic, Zoran
contents Speaker diarization (SD) struggles in real-world scenarios due to dynamic environments and unknown speaker counts. SD is rarely used alone and is often paired with automatic speech recognition (ASR), but non-modular methods that jointly train on domain-specific data have limited flexibility. Moreover, many applications require true speaker identities rather than SD's pseudo labels. We propose a training-free modular pipeline combining off-the-shelf SD, ASR, and a large language model (LLM) to determine who spoke, what was said, and who they are. Using structured LLM prompting on reconciled SD and ASR outputs, our method leverages semantic continuity in conversational context to refine low-confidence speaker labels and assigns role identities while correcting split speakers. On a real-world patient-clinician dataset, our approach achieves a 29.7% relative error reduction over baseline reconciled SD and ASR. It enhances diarization performance without additional training and delivers a complete pipeline for SD, ASR, and speaker identity detection in practical applications.
format Preprint
id arxiv_https___arxiv_org_abs_2509_15082
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle From Who Said What to Who They Are: Modular Training-free Identity-Aware LLM Refinement of Speaker Diarization
Chen, Yu-Wen
Ho, William
Topaz, Maxim
Hirschberg, Julia
Kostic, Zoran
Audio and Speech Processing
Speaker diarization (SD) struggles in real-world scenarios due to dynamic environments and unknown speaker counts. SD is rarely used alone and is often paired with automatic speech recognition (ASR), but non-modular methods that jointly train on domain-specific data have limited flexibility. Moreover, many applications require true speaker identities rather than SD's pseudo labels. We propose a training-free modular pipeline combining off-the-shelf SD, ASR, and a large language model (LLM) to determine who spoke, what was said, and who they are. Using structured LLM prompting on reconciled SD and ASR outputs, our method leverages semantic continuity in conversational context to refine low-confidence speaker labels and assigns role identities while correcting split speakers. On a real-world patient-clinician dataset, our approach achieves a 29.7% relative error reduction over baseline reconciled SD and ASR. It enhances diarization performance without additional training and delivers a complete pipeline for SD, ASR, and speaker identity detection in practical applications.
title From Who Said What to Who They Are: Modular Training-free Identity-Aware LLM Refinement of Speaker Diarization
topic Audio and Speech Processing
url https://arxiv.org/abs/2509.15082