%Aigaion2 BibTeX export from Idiap Publications
%Saturday 21 December 2024 04:40:58 PM

@INPROCEEDINGS{Kumar_ICASSP2024_2024,
         author = {Kumar, Shashi and Madikeri, Srikanth and Iuliia, Nigmatulina and Villatoro-Tello, Esa{\'{u}} and Motlicek, Petr and S, Karthik Pandia D and Dubagunta, S. Pavankumar and Ganapathiraju, Aravind},
       keywords = {F1 score, multitask learning, Speaker change detection, speaker turn detection, speech recognition},
       projects = {UNIPHORE, ELOQUENCE},
          month = apr,
          title = {Multitask Speech Recognition and Speaker Change Detection for Unknown Number of Speakers},
      booktitle = {Proceedings of the 49th IEEE International Conference on Acoustics, Speech, & Signal Processing (ICASSP) 2024},
           year = {2024},
          pages = {12592-12596},
      publisher = {IEEE},
       location = {Seoul, Republic of Korea},
           issn = {2379-190X},
           isbn = {979-8-3503-4485-1},
            url = {https://ieeexplore.ieee.org/document/10446130},
            doi = {https://doi.org/10.1109/ICASSP48485.2024.10446130},
       abstract = {Traditionally, automatic speech recognition (ASR) and speaker change detection (SCD) systems have been independently trained to generate comprehensive transcripts accompanied by speaker turns. Recently, joint training of ASR and SCD systems, by inserting speaker turn tokens in the ASR training text, has been shown to be successful. In this work, we present a multitask alternative to the joint training approach. Results obtained on the mix-headset audios of AMI corpus show that the proposed multitask training yields an absolute improvement of 1.8\% in coverage and purity based F1 score on SCD task without ASR degradation. We also examine the trade-offs between the ASR and SCD performance when trained using multitask criteria. Additionally, we validate the speaker change information in the embedding spaces obtained after different transformer layers of a self-supervised pre-trained model, such as XLSR-53, by integrating an SCD classifier at the output of specific transformer layers. Results reveal that the use of different embedding spaces from XLSR-53 model for multitask ASR and SCD is advantageous.}
}