When you train a model, you will have to select what kind of data you will train, "Conversations" or "General Text Data"

To model conversational data, we need a .csv file containing one message per row or integration with any available service, like Intercom

As columns we require:

  • Content: The text of the message.

  • Timestamp: When was the message sent?

  • Conversation Id: An identifier allowing us to group messages into conversations.

  • Person Type: Who sent the message? The client, the executive, a bot or the system?

The names and order of the columns are irrelevant, as we will later ask you to identify each of the above.

Additional columns in the .csv file are allowed for metadata.

Simply download this example file to get started!

Did this answer your question?