DiaCorrect: Error Correction Back-end For Speaker Diarization
This work addresses error reduction in speaker diarization, which is important for applications like transcription and meeting analysis, but it is incremental as it builds on existing error correction techniques from speech recognition.
The authors tackled the problem of speaker diarization errors by proposing DiaCorrect, an error correction framework that refines initial system outputs, resulting in effective improvements on 2-speaker telephony data.
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 transform-based 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.