Sure, I've got a sexting data set there's one on internet I can share the link but its like 10k lines of sexting convos , I've to augment that dataset using some algorithms they're known too , it's usually based on synonyms , for the conversation I've seen that a convo won't start with a show me your kitty directly , it'll be more of a get to know phase , at that point I had to train my model on labelisation too , I've had a normal convo data set and a sexting data set , each of them is labeled and fed to the model , so when I get a Hi , my ai answer to the Hi , when the convo start to heat up it automatically follow up in the horny phase