Datasets:

---

ZhangHanXD

/

AvaMERG

like 2

I found the conversation id "01835" in the train.json, whose response maybe belongs to the speaker rather than the listener? Besides, the utterance, the corresponding video and audio are seems like inconsistent. For example, the index 1 and index 2 are both the listener based on the train.json, but your multimodal file gives the listener's utterance 1 (index 1), speaker's utterance 2 (index 2).