- Introduction
- Setting up your account
- Balance
- Clusters
- Concept drift
- Coverage
- Datasets
- General fields
- Labels (predictions, confidence levels, label hierarchy, and label sentiment)
- Models
- Streams
- Model Rating
- Projects
- Precision
- Recall
- Annotated and unannotated messages
- Extraction Fields
- Sources
- Taxonomies
- Training
- True and false positive and negative predictions
- Validation
- Messages
- Access Control and Administration
- Manage sources and datasets
- Understanding the data structure and permissions
- Creating or deleting a data source in the GUI
- Uploading a CSV file into a source
- Preparing data for .CSV upload
- Creating a dataset
- Multilingual sources and datasets
- Enabling sentiment on a dataset
- Amending dataset settings
- Deleting a message
- Deleting a dataset
- Exporting a dataset
- Using Exchange integrations
- Model training and maintenance
- Understanding labels, general fields, and metadata
- Label hierarchy and best practices
- Comparing analytics and automation use cases
- Turning your objectives into labels
- Overview of the model training process
- Generative Annotation
- Dastaset status
- Model training and annotating best practice
- Training with label sentiment analysis enabled
- Training chat and calls data
- Understanding data requirements
- Train
- Introduction to Refine
- Precision and recall explained
- Precision and Recall
- How validation works
- Understanding and improving model performance
- Reasons for label 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
- Using general fields
- Generative extraction
- Using analytics and monitoring
- Automations and Communications Mining™
- Developer
- Exchange Integration with Azure service user
- Exchange Integration with Azure Application Authentication
- Exchange Integration with Azure Application Authentication and Graph
- Fetching data for Tableau with Python
- Elasticsearch integration
- Self-hosted Exchange integration
- UiPath® Automation Framework
- UiPath® Marketplace activities
- UiPath® official activities
- How machines learn to understand words: a guide to embeddings in NLP
- Prompt-based learning with Transformers
- Efficient Transformers II: knowledge distillation & fine-tuning
- Efficient Transformers I: attention mechanisms
- Deep hierarchical unsupervised intent modelling: getting value without training data
- Fixing annotating bias with Communications Mining™
- Active learning: better ML models in less time
- It's all in the numbers - assessing model performance with metrics
- Why model validation is important
- Comparing Communications Mining™ and Google AutoML for conversational data intelligence
- Licensing
- FAQs and more
Communications Mining user guide
Explore
The Explore page allows you to search, review, and filter a dataset to inspect and review individual messages and general fields. Select the Explore tab from the navigation bar to navigate the page:
By default, Explore presents the 20 most recent messages in a dataset in the Recent mode. You can select the dropdown mode selector to change this.
The different options you can select from the drop-down menu are:
- Recent – view 20 most recent messages.
- Shuffle – view 20 random messages.
- Teach – show 20 messages that the platform is unsure how to annotate.
- Low confidence – show 20 messages that are not well covered by informative label predictions.
- Rebalance – show 20 messages that are underrepresented by the training data in your dataset.
- Label – view 20 messages with the selected label assigned or predicted, which is the default mode when you select a label.
- Check label – view 20 messages that may have the selected label applied incorrectly.
- Missed label – view 20 messages that may be missing the selected label.
At the bottom of the page, you can select to move to the next page of 20 messages, or go back to a previous page.
This section explains the filters in Communications Mining™ and how to apply them.
The Filters bar allows you to find specific groups of messages, where you can filter:
- Specific date ranges, which allow you to select exact dates, or select from options such as the last week, month, 90 days, or year.
- Reviewed or unreviewed messages.
- Messages with positive or negative sentiment predictions, if sentiment is enabled on the dataset.
- Messages that have specific general fields predicted or assigned.
- Messages that include or exclude a specific label or a combination of predicted labels. For more details, check Advanced Prediction Filters.
Moreover, you can add any filter based on the metadata properties associated with your messages by selecting Add a new filter.
When you select Add a new filter, the drop-down menu shows a full list of all the available property filters.
These are naturally grouped by categories, and some are unique to the communication type in the dataset, for example, email.
The property categories that properties are grouped together in are:
- Source - only appears if there is more than one source in the dataset.
- Email - these are specific to individual emails, for example, who sent the email.
- Thread - these are email-specific and relate to the characteristics of email threads.
- Attachment - specific for messages, primarily emails, with specific attachment properties.
- User - all other metadata properties uploaded, not derived by the platform, with each message.
An icon indicates the property type for each property, whether it is a number or a string. For string user properties, the platform provides an example value when you hover.
When you add a filter for metadata fields with a string format, you can choose which to include or exclude in your selection, as shown in the following images:
If you add a filter for metadata fields with a number format, you can select minimum or maximum values, to create a range of your choice, as shown in the following image:
To remove a filter that you have applied, select the bin icon that appears when you hover over it with your mouse, or select Clear All to remove all filters applied.
You can use the label filter bar to filter messages that include or exclude specific labels predicted. You can do this either during model training, or when exploring and interpreting your data. For more details, check Advanced prediction filters.
You can use the following buttons in the Labels section to filter between showing all messages, to those that have had labels assigned to them, or those with prediction, which have not been reviewed. The icons appear as follows, and they change colour when selected:
| Select messages that have assigned labels. | |
| Select messages that have labels predicted. |
To deselect the filter, select the button again.
If you do not select any of these buttons, but filter to a label, the platform will filter to all messages that either have the label pinned or predicted, starting with the reviewed messages first.
The label filter bar and the + Add label filter allow you to add complex combinations of inclusion and exclusion filters, for example, show me messages with X and Y predicted, but not Z. For more details on how to use these filters, check Advanced prediction filters.
Red dial training indicator
- The red dial training indicator shows up for some labels and highlights the ones that require more training examples for the platform to accurately evaluate the performance of the label. For more details, check Reviewing messages.
- The completeness of the circle indicates how many more examples are needed. The larger the red section, the more examples are required.
- Once you have 25 annotated examples, the red circle disappears, depending on the complexity of the label. However, you may need more examples to get accurate predictions.
- You should review messages to find more training examples.
For datasets containing emails, these are displayed, showing the email that matches the selected sort order, for example, Teach Label, Missed Label, and so on, but with easy access to the other emails that are in the same email thread.
In the following example, you can notice that the sorted email is in a thread of three emails, and this is the third email in the thread.
You can expand out the email thread to show partial views of the other emails in the thread by selecting the bi-directional arrow icon below the subject:
If you select again on any of the partially expanded emails, they will expand fully, as the original sorted email:
