- Getting Started
- Administration
- Manage Sources and Datasets
- Understanding the data structure and permissions
- Create a data source in the GUI
- Uploading a CSV file into a source
- Create a new dataset
- Multilingual sources and datasets
- Enabling sentiment on a dataset
- Amend a dataset's settings
- Delete messages via the UI
- Delete a dataset
- Export a dataset
- Using Exchange Integrations
- Preparing Data for .CSV Upload
- Model Training and Maintenance
- Understanding labels, entities and metadata
- Label hierarchy and best practice
- Defining your taxonomy objectives
- Analytics vs. automation use cases
- Turning your objectives into labels
- Building your taxonomy structure
- Taxonomy design best practice
- Importing your taxonomy
- Overview of the model training process
- Generative Annotation (NEW)
- Understanding the status of your dataset
- Model training and labelling best practice
- Training with label sentiment analysis enabled
- Train
- Introduction to 'Refine'
- Precision and recall explained
- Precision and recall
- How does Validation work?
- Understanding and improving model performance
- Why might a label have low average precision?
- Training using 'Check label' and 'Missed label'
- Training using Teach label (Refine)
- Training using Search (Refine)
- Understanding and increasing coverage
- Improving Balance and using 'Rebalance'
- When to stop training your model
- Applying labels
- Reviewing messages
- Searching for messages
- Label editing
- Using Analytics & Monitoring
- Automations and Communications Mining
- FAQs and More
Reviewing messages
User permissions required: ‘View Sources’ AND ‘Review and label’.
Overview
Reviewing unreviewed messages and accepting or rejecting the platform's predicted labels and entities further trains the model and its accuracy.
You can review unreviewed messages in most of the training modes in Explore and in Discover:
- Cluster (Discover)
- Search (Discover & Explore)
- Recent (Explore)
- Shuffle mode (Explore)
- Label mode (Explore)
- Teach (Explore)
- Low Confidence (Explore)
Make sure to apply all of the relevant labels in your taxonomy to each message. When you review a message, not only do you teach the model which labels apply, but also which labels don’t. If you don’t apply all relevant labels, you send a negative training signal to the model, which will affect its performance.
The opacity of a label indicates the confidence of the platform's prediction of that label, with higher opacity indicating higher confidence.
Hovering your cursor over the label opens a modal showing the confidence with which the model has predicted the label and, if sentiment analysis is enabled, the net sentiment.
- Clicking on the label, or the sentiment indicator (if sentiment analysis is enabled) pins the label to the message, i.e. it confirms the model’s prediction of that label
- If you want to change the sentiment of the predicted label, click the face image that appears when you hover over the message
- If the prediction is wrong, add the correct one - this effectively dismisses the incorrect predictions
Hovering your cursor over the entity opens a modal showing the confidence with which the model has predicted the entity.
Accept/reject an entity
- Clicking 'Confirm' on the entity (or clicking the hotkey - which is '1' to confirm an entity), pins the entity to the message, i.e. it confirms the model’s prediction of that label
- Clicking the 'Dismiss' on the entity (or clicking the hotkey - which is '2' to dismiss an entity), tells the platform that the entity predicted is incorrect
- Clicking the change entity button allows us to assign a different entity, if the entity predicted is incorrect
- In the example above, clicking this button will display the other entities in our dataset that we can assign
- In this case, we can change the entity type from 'Cancellation Date' to 'Policy Start' on this dropdown, which will assign this entity
