Publication detail
Diacorrect: Error Correction Back-End for Speaker Diarization
HAN, J. LANDINI, F. ROHDIN, J. DIEZ SÁNCHEZ, M. BURGET, L. CAO, Y. LU, H. ČERNOCKÝ, J.
Original Title
Diacorrect: Error Correction Back-End for Speaker Diarization
Type
conference paper
Language
English
Original Abstract
In this work, we propose an error correction framework, named DiaCorrect, to refine the output of a diarization system in a simple yet effective way. This method is inspired by error correction techniques in automatic speech recognition. Our model consists of two parallel convolutional encoders and a transformerbased decoder. By exploiting the interactions between the input recording and the initial system's outputs, DiaCorrect can automatically correct the initial speaker activities to minimize the diarization errors. Experiments on 2-speaker telephony data show that the proposed DiaCorrect can effectively improve the initial model's results. Our source code is publicly available at https://github.com/BUTSpeechFIT/diacorrect.
Keywords
Speaker diarization, error correction, conversational telephone speech
Authors
HAN, J.; LANDINI, F.; ROHDIN, J.; DIEZ SÁNCHEZ, M.; BURGET, L.; CAO, Y.; LU, H.; ČERNOCKÝ, J.
Released
14. 4. 2024
Publisher
IEEE Signal Processing Society
Location
Seoul
ISBN
979-8-3503-4485-1
Book
ICASSP 2024 - 2024 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)
Pages from
11181
Pages to
11185
Pages count
5
URL
BibTex
@inproceedings{BUT189697,
author="HAN, J. and LANDINI, F. and ROHDIN, J. and DIEZ SÁNCHEZ, M. and BURGET, L. and CAO, Y. and LU, H. and ČERNOCKÝ, J.",
title="Diacorrect: Error Correction Back-End for Speaker Diarization",
booktitle="ICASSP 2024 - 2024 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)",
year="2024",
pages="11181--11185",
publisher="IEEE Signal Processing Society",
address="Seoul",
doi="10.1109/ICASSP48485.2024.10446968",
isbn="979-8-3503-4485-1",
url="https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10446968"
}
Documents