Once more Diarization: Improving meeting transcription systems through segment-level speaker reassignment - View it on GitHub