Working with Conversation Data#
HumanFirst allows you to upload data as individual data points or as utterances (conversational turns) within a conversation. When uploading a CSV, the default option is to treat each row as an individual data point. Select the conversations option if your data represents multi-turn exchanges.
Required Columns for Conversation Data#
Uploading data as a conversation or set of conversations requires the following columns in your CSV:
Conversation ID
An alphanumeric ID shared by each utterance within the same conversation. All rows with the same Conversation ID will be grouped together as a single conversation.
Date and Timestamp
A date and timestamp for each utterance. Accepted formats: Unix milliseconds, Unix seconds, RFC 3339, YY-MM-DD HH:MM:SS, MM/DD/YY HH:MM, or MM-DD-YY HH:MM:SS.
Source
A column denoting who is speaking. Conversation format supports up to 2 speakers.
Client Value
The value used to identify customer utterances. These will be displayed in the data view as individual, searchable, and clusterable data points.
Agent Value
The value used to identify agent utterances. These are included in the conversation transcript but are not displayed as searchable or clusterable items.