guide
- Get started
- Label Studio features
- Billing & Usage
- Release notes
Security and Privacy
- Secure Label Studio
Install and Upgrade
- Install and upgrade Label Studio
- Database setup
- Start Label Studio
- Troubleshooting
Deploy and Install Enterprise
- Install Label Studio Enterprise
- Set up persistent storage
- Set up an ingress controller
- Install with Docker
- Deploy on Kubernetes
- Install on airgapped server
- Install on Amazon EKS
- Available Helm values
Manage Users
- Set up user accounts
- Manage access
- Set up authentication
- Set up SCIM2
Import Data
- Import data
- Import pre-annotations
- Cloud storage setup
Labeling and Projects
- Project setup
- Manage data
- Set up your labeling interface
- Label and annotate data
- Comments and notifications
Manage Annotations
- Review annotations
- Annotation statistics
- Custom agreement metric
- Export annotations
Machine Learning Setup
- Machine learning integration
- Write your own ML backend
- ML Examples and Tutorials
- Active learning loop
- Troubleshoot machine learning
Integrations
- Webhook Setup
- Webhooks Event Reference
- Custom Webhooks
- Python SDK Tutorial
- Backend API
Advanced Development
- Frontend library
- Frontend reference
- Update scripts and API calls
Comments and notifications
Annotators, reviewers, and administrators can use comments to do the following:
Users can discuss task issues and other problems during labeling and reviewing processes using Comments and Notifications in Label Studio Enterprise.
Workflow overviews
The following two workflows are flexible and can be combined. For example, in Figure 1A, Annotators don’t have access to the Data Manager and they would only use the Label Stream. They can write comments there, resolve them, and see notifications. Meanwhile, other roles like Managers or Administrators can filter tasks by “Unresolved comments” in the Data Manager and use the second workflow (see Figure 1B).
Figure 1A. The workflow based on notifications
Figure 2B. The workflow based on the Data Manager
Notification workflow
This approach is centered on the notification panel (see figure 1A). The entry point here is the notification link, it allows to go into editing or review mode (depending on your role). Note that label or review streams are deactivated there, only a task from a notification link is shown.
Annotators
Write comments
Read other’s comments
Annotators will see notifications from reviewers and other users when another user’s comment is added to the annotator draft or annotation
info
An annotator must be added to the project or workspace to see associated comments.
Reviewers and other higher roles
Write comments
Reviewers (or other higher roles) can write comments during the review stream. It is especially useful when a reviewer rejects an annotation.
Read comments
info
A reviewer must be added to the project or workspace to see associated comments.
Submitted annotations vs postponed drafts
The standard labeling workflow is to submit an annotation. This action affects statistics, agreement, and annotator performance scores. In some cases where annotators want to avoid submitting (e.g. while they wait for their comment to be resolved), they can use the postpone draft functionality by clicking blue arrow button in the labeling stream:
It’s helpful to combine postponed drafts with comments, because users can clarify some questions without affecting a dataset labeling state. Annotators can access their postponed drafts in the following ways:
Note: If the user refreshes the page, the navigation history will be lost. This inconvenience will be fixed soon in upcoming releases.
Data manager workflow
You can add comments to tasks independent of annotation and action such as Skip or Accept.
In the Label Studio UI, navigate to your Data Manager.
Click on the Columns drop-down to filter the columns based on the following:
Figure 2: Columns drop-down.
Figure 3: Commented at.
Now, you can also filter by the number of comments and comments added or by the authors using the Filter drop-down located next to the Columns drop-down. You can add more filters to display tasks based on your filtering requirements.
Figure 4: Filter to comment by author.
To resolve comments, click on a task and on the right side of the window you will see the Comments section where you can add a comment or click the three dots (…) to resolve or unresolve comments.
Figure 5: Resolve comments.
When you click Resolve, you will see a green color check mark next to your comment. This represents a resolved comment.
Figure 6: Green color check mark for resolved comments.
To unresolve comments, click the three dots (…) >> Unresolve button.
Figure 7: Unresolved comments
There is a special pink icon that appears on comments attached to annotations or drafts in the dropdown menu that indicates unresolved comments:
Comments
The Comments feature allows Annotators Reviewers and other higher roles to communicate and discuss queries on Label Studio Enterprise. Users will be notified of comments if they are associated with the task, and new comments will show up in the Notifications panel for a quick reference. Users can also filter and sort tasks by Comments status in the Data Manager view, such as finding tasks with unresolved comments from a particular user.
Roles and Capabilities
All users
All users with access to Label Studio UI can do the following:
Annotators
As an Annotator, you can do the following:
Reviewers and higher roles
As a Reviewer, you can do the following:
note
Permissions
Table 1 displays the three key user actions on the Comments feature.
Table 1: User permissions.
Project settings
You can make it mandatory to write comments when annotators click Skip on annotations. You can find it in Project Settings on the Annotation tab:
A similar option is available for reviewers. You can find it in Project Settings on the Review tab:
Notifications
To see the notifications, navigate and click the user profile icon on the top-right of the Label Studio UI. The Notifications pane opens with a bell icon next to it. The blue color bell icon indicates unread notifications and the grey color bell icon indicates read notifications.
Figure 8: Notification panel.
info
The read notification has a grey color background which means a user has already read this notification.
If you found an error, you can file an issue on GitHub!