- 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
Using reports
The Reports page serves as the hub for in-platform reporting on your dataset, offering valuable insights and statistics. These reports are highly customizable, allowing you to focus on the views that fit your use cases the most. Accessible through the navigation bar, Reports allows you to better understand your data.
Depending on your data type, the tabs in Reports may vary. Each tab is designed to address different reporting needs:
- Dashboard - Create custom dashboard views using data from other tabs.
- Label Summary - Provides high-level summary statistics for labels.
- Trends - Displays charts for message volume, label volume, and sentiment over a selected time period.
- Segments - Offers charts comparing label volumes to message metadata fields, such as sender domain.
- Threads - Shows charts of thread volumes and label volumes within a thread, accessible when you apply the Thread.
- Comparison - Allows you to compare different cohorts of data against each other.
The Dashboard page is where you can create fully customizable dashboards containing all of the charts and visuals in the platform that are most useful or relevant to them for this dataset.
Dashboards are specific to the dataset itself. Each user has a default dashboard automatically created, where you can create new dashboards, and delete and edit existing dashboards as well. You can rename the dashboard titles to suit your use case.
The Dashboard page is particularly useful when you have a live integration set up for the sources in your dataset, and you can periodically check in to observe how things have changed over time.
As a default, each dashboard has a few example visuals already added, however, you have complete control over what gets shown on each dashboard.
You can customize a dashboard by adding or removing charts with your choice of filters applied, or specific data visuals from the Dataset settings page. You can also edit the filters applied to charts in the dashboard from within the dashboard itself.
You can also rename the dashboard by selecting the title and entering the new name.
As you can edit multiple dashboards, you first need to ensure that you have the intended dashboard selected.
To select the right dashboard, go to the Dashboard page, select the dropdown menu from the page, and select the dashboard you want to update.
When you go to Reports to add charts or visuals, they are added to the dashboard you just selected on the Dashboard page.
To rename any dashboard, select the dashboard title, and enter the new name. This will then be reflected in the main dropdown menu as well.
As a default, a dashboard contains the default visuals as shown in the following image. You can also remove all of them, if you wish.
When you add a chart to a dashboard, the platform will intuitively try to fit it in the most obvious space available. If there is space to accommodate the default chart size, the platform will add the chart as close to the top of the page as it can.
To add a chart, proceed as follows:
- Ensure you have the intended dashboard selected in the Dashboard page as shown in the previous image.
- Navigate to the Reports page through the dataset navigation bar, or where Reports is shown on the empty Dashboard page as previously shown.
- Apply the relevant filters to find the chart that you want. For more details, check Applying filters in Reports.
- Select the plus + icon that appears on the chart when you hover over it with your mouse.
- This will add the chart to the selected dashboard with the filters and chart type selected at that point in time
- Repeat this process from the Reports page until you have added all of the chart views that you want in your dashboard
When you add a chart from Reports with filters applied, you will notice the filter icon
as shown in the following image. Select it
to reveal which filters are applied, and the icon will change its color
.
To add or remove dashboard cards, select the image icon , and
then select which cards you want to display:
The Recent Messages card also has some additional options available in the dashboard. Select the ellipsis on this card, and select one of the following options:
- Hide labels - Hide the predicted labels for the recent message.
- Hide user properties - Hide the user properties, that is, the metadata of the recent message.
- Show single message - Toggle between showing a single message or multiple messages. This will increase the number shown from 1 to 5.
- Edit card filter - Edit the filters applied to the card.
- Remove card - Remove chart from the dashboard, which is a standard option for all charts.
For any chart that you have added to a dashboard from the Reports page, you can select the
chart icon on a chart, which redirects you to a view
of that chart with its applied filters in the Reports page.
From there you can download an image of the chart or the underlying data, or you can select the underlying messages in Explore.
Once you create a dashboard, you can edit or add further filters to it. You can apply a Global filter to the dashboard, augmenting all existing dashboards cards with a new filter. You can also edit individual card filters , if you want to update the existing filter without having to create a new card from scratch in Reports.
- Navigate to the Dashboards page in Reports.
- Select the main filter icon
, which reveals the Global filters side panel.
- Configure your filters in the side panel.
To edit the existing filter on a card, select the filter icon on each card.
The charts and visuals you add to a dashboard can be rearranged by selecting the top of the chart and dragging and dropping them wherever you want on the page.
The Dashboard page layout is defined by a column system, between 1 and 6 columns wide, depending on the size of your screen. A smaller laptop screen may only fit 3 columns, whilst a much larger screen or TV could fit 6.
This means that, depending on how you resize them, the width of any of your charts will snap to be any number between 1 and 6 columns wide. If your screen only accommodates 3 columns, you could have a chart that is 1 column wide side by side with a chart that is 2 columns wide, or 3 that are 1 column wide, and so on.
Vertically, the charts can be whatever height you want.
- When you set up the arrangement of your dashboard, this will be for the specific screen width you are using at the time, such as 3 columns.
- If you regularly switch between 2 differently sized screens, you should set an arrangement for each of those screen sizes.
- The platform will then automatically revert between the two depending on which screen you’re using at the time.
- The first time you switch to a new screen size, the dashboard will arrange all of the charts in a vertical list as default.
- Navigate to the Dashboards page in Reports.
- Select the ellipsis on a chart.
- Select Remove Card.
You can always readd the chart later from the Reports page.
To change the size of your charts, select on the arrow icon on a chart, and drag it to the column width or any height that you want.
The area surrounding the chart will be highlighted blue until the chart is the size that you want, and you release the mouse.
The blue area indicates the width that the chart will snap to, based on the number of available columns on your screen. Remember, your dashboard will be between one and six columns wide, depending on the width of your screen.
Under the Label Summary tab, there are a number of different charts and tables available:
- Tree map
- Top X highest volume leaf labels
- Top X highest volume leaf labels
The tree map is a way of visually showing your taxonomy hierarchy, while also displaying quantities for each label via the area size.
Each root label is assigned a rectangular area with their leaf labels, that is, their sub-labels, displayed as rectangles nested inside of it. When a label count is assigned to a label, its area is displayed in proportion to that label count number and to the other quantities within the same parent category.
When you select a label name it will take you down a level to the leaf labels that are nested within. To get back to the top level you select the bar with the root label name:
This chart shows by default the top 10 highest volume leaf labels over the whole dataset. This is also calculated on a sum of probabilities of the predicted labels.
When sentiment is applied the chart will show both positive and negative labels. The top 10 is a split between the top 5 positive labels, and the top 5 unique negative labels.
This means that if the top negative label is already included in the top 5 positive labels, then it takes the next highest one as the top negative label and so on. This way the chart is always showing 10 unique labels.
The top X highest volume leaf labels table shows the same as the chart but in table format, with each label and corresponding label count. It defaults to show 10 labels, but you can adjust it to show the top 5 or 20 labels if preferred.
This is also calculated on a sum of probabilities of the predicted labels.
When sentiment is applied the table shows a split of the top 10 positive labels and the top 10 unique negative labels.
Under the Trends tab, there are a number of different charts and tables available.
- Message volume over time
- Label volume and sentiment trends
- Label trends
This page is particularly useful for reviewing how labels have changed over a given time period.
You can adjust the time sequence, that is, daily, weekly, monthly, annually, of the chart period using the dropdown from the Reports page, or filter the chart to a specific time series using the filter bar or by selecting an area on the chart. For more details, check Applying filters in Reports.
The label volume trends chart shows the trend of the top 6 highest volume labels over the whole dataset as a default. As with the tree map, the percentage numbers are calculated as a sum of probabilities of predicted labels.
You can also filter the time period on this chart by using the dropdown menu or selecting the time period in the chart.
If you want to plot specific labels on this or other label-specific charts, select those labels through the Charted labels dropdown from the page.
If sentiment analysis is enabled on your dataset, there will also be a label sentiment trend over time chart available in Trends.
The label trends table shows you how the top 10 labels for a given time period perform compared to the previous period and their change in rank.
The time series in scope is set by the user in the filter bar, and the time sequencing of the chart, that is, daily, weekly, and so on, in the top right dropdown menu.
For the time period selected, the table shows the top 10 labels based on the sum of probabilities of predicted labels. It also highlights how those labels have changed in rank from the previous period.
If sentiment analysis is enabled, it totals covering both positive and negative sentiment labels.
By adjusting the time sequence filter, this is a useful way of seeing if there were spikes or dips in volumes in a certain label, or group of labels, in a given period.
For each time sequence it will show the following period:
- Daily - The top 10 labels yesterday compared with the previous 7 days.
- Weekly - The top 10 labels from the previous week compared with the previous 4 weeks.
- Monthly - The top 10 labels from the previous month, compared with the previous 12 months.
- Yearly - The top 10 labels from the last year, counting back from yesterday, with the previous year.
The Segments tab shows charts of label volumes split by message metadata category, that is, by user properties, as well as a chart with the associated top 6 labels for each of those categories.
When data is uploaded to the platform it normally contains metadata attached to each message. Some examples for different types of data include:
- Email – number of emails in chain, recipient name, inbound/outbound, emails in thread, mailbox name, recipient domain.
- Feedback – satisfaction rating score, product code, location, business unit, NPS score.
- Surveys – age, gender, year, question, business area, location, status, country.
- Chat – agent, chat purpose, client business.
- Call recordings - agent, chat purpose, client business.
If the metadata can be grouped into categories, such as distinct sender domains, then a chart can be created showing the volume of messages, split by category, such as each sender domain.
The platform loads certain charts as default, some are hidden and listed at the top of the page. Select them to unhide them. For more details, check Reports.
The chart will limit the categories on the chart to the top 20. To the right of this chart will show the top 6 labels for each of those categories.
The Label Volume chart shows the top 6 labels for each user property, that is, each category. For each label in that category it shows the percentage messages in that category that the label is predicted for. This is also a sum of probabilities of predicted labels.
customer.com
.
customer.com
are predicted to have
the label Claim.
Under the Segments tab you cannot create a chart from a category that is number based, with the exception of NPS score. If the metadata has numbers that have been categorised, say feedback scores of 1 to 5 or years, then it can be uploaded as string values and this would be shown in the Segments charts.
A chart is not created where there are numeric properties, such as monetary amounts, percentages, number of participants or recipients of an email, and so on. However, you can use these as filters.
This can be done by using the filters from the side panel, and when applied, the charts in segment will be filtered on these criteria:
To understand more about the user property filters, check Explore.
To find out about advanced prediction filters for labels, check Advanced prediction filters.
Under the Threads tab of the Reports page, there are a number of different charts available if you have threaded messages, such as email chains, chats, and so on within your dataset:
- Number of Messages
- Thread Duration
- Number of Participants
- Response Time
This page is particularly useful for reviewing some of the thread-specific metrics.
At the top of the filter bar, you can toggle between Messages and Threads. This allows you to apply analytics at the level of conversations, instead of individual messages.
The threads filter is applicable for longer-form conversations like email threads, phone calls and live chats. Once the Threads filter is selected, the fifth Threads tab will become available.
The platform will display the threaded volume by number of messages, as well as the top label volumes by number of messages.
The number of messages is calculated by taking a count of the number of messages that are part of the same thread.
The platform will display the thread volume by duration, as well as the top label volumes by duration.
The thread duration is the time between the first and the last message that was sent in the thread.
The platform will display the thread volume by number of participants, as well as the label volume by number of participants.
The number of participants is calculated by adding up all parties To, From, Cc, and Bcc, if the mailboxes attached as sources have been copied in.
The platform will display the thread volume by response time, as well as the label volume by response time.
The thread response time is calculated by checking how long it takes for the original sender of the message to receive a response from someone other than the original sender.
You can update the Comparison page to compare groups of messages or threads and conduct A or B testing. This functionality has many useful applications, such as testing the impact of a new marketing initiative or change programme over a time series, or between different regions or customer groups.
- Navigate to the Comparison page in the Reports tab on the navigation bar.
- Toggle between messages and threads, depending on the analytics you want to extract. You can compare individual messages, or you can compare at a thread level, such as for email threads, phone calls, and live chats.
- Filter the two different groups, A and B, using the filter bar from the side
panel.
Note: The filter bar works in the same way as the filters in the Explore and Reports pages, except that you cannot filter by general fields or labels, only metadata, including dates.
You can also copy the filters from one group to another, before tweaking one group further, to save selecting similar filters twice.
Once you have set the filters, you can view a comparison between the two groups, which shows per label:
- Total message count - across all messages in both Group A and Group B.
- Proportion (%) – the percentage of the messages in both Group A and Group B with this label.
- Differences (%) – the proportional differences for each label between Group A and Group B, calculated as Group B minus Group A.
The comparison table is sorted as default in descending order by message count in Group A. You can sort the table in ascending or descending order for any of the value columns by selecting the column name, and then the arrow that appears. To switch between the two, select it again.
Use the label filter bar to further filter what you are shown in the Compare page. Enter a root or leaf label that you want to display in the label column, and all other labels will be filtered out.
You can download as a CSV file the comparison table that shows in the Compare page also. The CSV file contains the details of the filters that were applied to create Group A and Group B.
To download the taxonomy and the group analysis, select the download icon from the main Comparison page.
- Using dashboards
- Introduction
- Selecting a dashboard
- Renaming the dashboard
- Adding charts
- Filtering the dashboard
- Rearranging your dashboard
- Removing a chart from your dashboard
- Resizing a chart on your dashboard
- Using Label Summary
- Tree map
- Chat for the top X highest volume leaf labels
- Table for the top X highest volume leaf labels
- Using Trends
- Message volume over time
- Label volume and sentiment trends
- Label trends
- Using Segments
- Overview
- Enabling and understanding user property charts
- Filtering user property charts
- Using Threads
- Enabling the Threads tab
- Number of Messages
- Thread Duration
- Number of Participants
- Thread Response Time
- Using Comparison
- Comparing different groups
- Sorting the comparison table
- Label filter
- Downloading the comparison table