shared catalogs?

[dcsan]

Are there any shared archives of threaded chat conversations? I'm trying to build a bot based on purely machine learning stuff, but my own logs aren't extensive enough.

I looked through some of the https://parl.ai/ corpus but there isn't much in the way of threaded conversation material.

any recommended sources would be appreciated.

[MasterScrat]

This is indeed a challenge, yes.

I had started gathering some celebrity interviews here: https://github.com/MasterScrat/interview-transcripts

But I gave up on it as it's not enough data (you'll have to look back a few commit, the interviews were previously in WhatsApp format.)

I also asked on Reddit but not much success: https://www.reddit.com/r/MachineLearning/comments/eef62v/p_looking_for_interview_transcripts_to_train_gpt2/

[mar-muel]

The Personachat dataset doesn't work for you? Apart from the conversations being fairly short, you get pretty good results for small talk behavior.