我又来了, 还是中文断句继续优化的样本

The red part indicates that the subtitle gap is less than 100ms, which is not a good reason to force a break. It does not recognize "Hanna*Arendt" and "Olivier*Grenier", which is just fine. Are these places just at the boundary of the batch or segment "feeding"? Or, in this special transcription mode, should there not be a way to avoid the simple fixed-length hard segmentation? After all, there must be a way to avoid this crude method of fixed-length hard segmentation, right?