I have a database full of conversations (chats between human-human), and I am wondering how can I load and annotate them using Rasa/Rasa X.
At the moment, I built a bot with minimum data (using 100 conversations manually annotated and load in Rasa). Now I am interested to add more data in batch format and annotate them to improve my bot.
Does anyone have any related experience to help me out?
Thanks,
Thanks for the fast reply. No my data came from real conversations collected from human-human interactions, they are not in āRasa Trackerā format.
I have built a MV bot with minimum data, I would like to load bunch of data/conversations in Rasa X at once (and start annotating them) instead of interacting one by one (bot- user, bot-user).
@Tobias_Wochinger: Thanks for the help. So I will need to convert conversations into āUser sent messageā and āBot responded messageā events, right?
For earlier one, I will need to run MV bot over user inputs to get intent classification output, right?
For āBot responded messageā event, I just need to have response texts?
After creating those events what will be the next steps?
Now I am more concerned about steps after creating āUser sent messageā and āBot responded messageā events. How can we move to create stories somehow automatically from conversation logs.
I donāt see any other solution than writing a tool which converts it and then probably do some manual work to polish the automatically created stories.