ixp
latest
false
- 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
Training chat and calls data
Important :
Communications Mining is now part of UiPath IXP. Check the Introduction in the Overview Guide for more details.

Communications Mining user guide
Last updated Aug 1, 2025
Training chat and calls data
User permissions required: ‘View Sources’ AND ‘Review and annotate’.
Note: Users will be able to see chat and calls data if they have ‘View Sources’ AND see labels if they have ‘View labels’ permissions,
but they will require the ‘Review and annotate’ permission in order to actually apply labels.
Chat/calls data are commonly trained for analytics and monitoring-based use cases to gain a detailed understanding of the processes, issues, and sentiments within a conversation.
Some examples of questions you can answer for these communication types:
- How many conversations start with a customer asking us about a topic, a complaint, etc.?
- What are the top topics customers are contacting us about?
- How long does it take to resolve a conversation about a given topic?
- What is the quality of service that agents are providing for our customers?
- What is the sentiment when a certain topic is mentioned?
A chat/call thread

Layout explained:
- This is used to indicate that a message has been marked as uninformative
- This indicates that a label has been added onto a message
- This allows a user to mark a message as uninformative
- This allows a user to add a label onto a message
- This allows a user to play back an audio recording, control the speed/volume, or download a call.
Note: If you have sentiment analysis enabled on your chat/calls data, the differences when annotating are the same as annotating
with sentiment for other communications channels (i.e. - assigning a sentiment each time you assign a label, using neutral
label names, etc.). See here for more details on annotating with sentiment analysis.
Training chat/calls data is very similar to training other message types, where a user would go through the Discover, Explore, Refine phases to train their model further.
The key distinctions are:
- Thread layout - With chat/calls data, messages between all parties in a given conversation are automatically compiled into a single thread view, but labels are still assigned to individual messages (i.e. - turns in the conversation).
- Uninformative messages - A message in a chat/call can be marked as 'uninformative' if it does not add context or value to the given conversation.
By marking a message as uninformative, you are teaching the model that none of the labels are applicable, and therefore the
model will learn that similar messages should not be expected to have label predictions.
Note: When applying labels to a message ('message A'), the model will automatically mark the previous message Message B as uninformative if no labels are applied to it. It's therefore important to read the previous message and apply labels to it if relevant. This feature helps to build up the necessary training data for 'Uninformative', without too much additional annotating.
- Coverage - When assessing coverage for chat/calls data, in addition to assessing the proportion of messages covered by informative (i.e - meaningful) label predictions, it also incorporates the proportion of messages that are predicted to be uninformative. For more information on how coverage is determined, select here.
Validation factor card for coverage for a chat or calls dataset