# Meeting scenario optimization

This is a speaker segmentation model based on powerset encoding, capable of processing 10-second audio clips and identifying multiple speakers and their overlapping situations.

This is a speaker segmentation model based on powerset multi-class cross-entropy loss, capable of processing 10-second mono audio and outputting speaker segmentation results.

A Chinese speech recognition model fine-tuned based on whisper-large-v3-zh, significantly improving punctuation recognition while maintaining original performance standards.