Investigating Confidence Estimation Measures for Speaker Diarization
This work addresses errors in speaker diarization for applications like speech recognition, but it is incremental as it focuses on improving confidence estimation rather than a fundamental breakthrough.
The paper tackled the problem of speaker diarization errors propagating to downstream systems by investigating methods for generating segment-level confidence scores, finding that the best methods could isolate about 30% of errors within the lowest 10% of confidence scores.
Speaker diarization systems segment a conversation recording based on the speakers' identity. Such systems can misclassify the speaker of a portion of audio due to a variety of factors, such as speech pattern variation, background noise, and overlapping speech. These errors propagate to, and can adversely affect, downstream systems that rely on the speaker's identity, such as speaker-adapted speech recognition. One of the ways to mitigate these errors is to provide segment-level diarization confidence scores to downstream systems. In this work, we investigate multiple methods for generating diarization confidence scores, including those derived from the original diarization system and those derived from an external model. Our experiments across multiple datasets and diarization systems demonstrate that the most competitive confidence score methods can isolate ~30% of the diarization errors within segments with the lowest ~10% of confidence scores.