Conversational

Conversational labeling allows you to label spans in a text document, much like Span labeling.

The key difference is the UI: conversational labeling uses a chat-style interface with bubbles and avatars for each line or "message," making it easier to label conversational data in a familiar interface. A standout feature is that Conversational projects allow you to label entire messages, which is ideal for classifying conversations.

Additionally, the same label sets used in Span labeling projects are compatible with Conversational projects.

Conversational Overview

How to create a conversational project

1. In Step 1 of the Project Creation Wizard, upload a file with the type that is supported for Conversational Labeling. You can use the sample JSON file below to setup the project.

4KB

Datasaur Sample - Conversational.json

1. In Step 3 of the Project Creation Wizard, you can set up your label set either through the UI or by uploading a label set file. You can use the sample CSV file below to quickly configure your label set classes.

94B

Datasaur Sample - Conversational Label Set.csv

For more detailed information on Conversational projects, please refer to this page.

Additional settings in conversational project

Conversational projects also support the additional settings that are available for Span labeling projects.

• Limit selection to a span of 1 token.

• Spans should have at most one label.

• Allow arrows to be drawn between labels.

• Default text selection, between token or character selection.