The cross dataset should be RAMC(ASR) dataset mixed with LibriTTS, but not RAMC(ASR) mixed with AISHELL3?

[simpleishappy]

The dataset provided seams wrong. There no scripts to generate the synthetic conversation dataset with two speakers that was created using the statistical distribution for gaps and overlaps described in prior psycho-linguistics work.