In today’s fast-paced business world, data is the lifeblood of organizations seeking to gain a competitive edge. Understanding the vast amounts of data generated within a company is crucial for making informed decisions. UiPath Communication Mining is a powerful tool that enables businesses to analyze and extract valuable insights from their communication data, such as emails, chat messages, and more.
Implementing UiPath Communication Mining:
In this step-by-step guide, we will walk you through the process of implementing UiPath Communication Mining, helping you unlock its full potential and harness the power of your data.
As discussed in the previous article, UiPath Communication Mining is an innovative technology that leverages artificial intelligence and machine learning algorithms to analyze unstructured communication data, including text messages, emails, audio transcripts, and social media interactions.
Uipath Communication Mining Key Concepts:
Before delving into the implementation process, let’s explore some of the key concepts of UiPath Communication Mining and How It Works…
🏢 Projects: Confined workspaces for managing specific data sets, linked to sources and datasets. Users need appropriate permissions to access data in a project, facilitating collaboration and analysis within teams or business units.
📊 Datasets: A dataset allows the labelling of one or more sources to build a model. Sources can be included in multiple datasets. The labels in a dataset form a taxonomy, which can vary based on the use case, such as analytics or automation. For example, a dataset could encompass sales conversations to monitor customer experiences.
🔍 Source: Data is organized into sources, corresponding to channels like email, surveys, or customer reviews. Multiple sources can be combined to build a model, promoting better organization. Raw collections of verbatims (freeform text communications) grouped together, including responses from surveys, emails, chat transcripts, and call records. They are incorporated into datasets to structure and interpret the verbatims.
🗒️ Verbatims and Metadata: Each verbatim contains associated metadata with structured data points providing additional information, including a timestamp for when it was created.
🗂️ Clusters: Groups of similar verbatims organized using unsupervised learning techniques, essential for subsequent stages of model training. The platform presents 30 clusters at a time on the Discover page, each containing 12 verbatims.
🏷️ Labels: Structured summaries of intents or concepts expressed within verbatims. Verbatims can have multiple labels, not limited to exclusive classifications. Labels are applied during model training and returned as predictions. They have confidence scores indicating the model’s likelihood of the prediction. A threshold can be set to convert the prediction into a “Yes/No” answer.
💬 Comment: Each piece of text communication within sources is represented as a comment with an ID, timestamp, text body, and additional fields specific to the type of data. For example, emails have fields like “from” and “to,” while customer reviews have the review author.
🧠 Model: The model is continuously updated as users label more data. When querying the model for predictions, a specific model version number needs to be specified for consistent results.
Setting Up UiPath Communication Mining Account:
To access Communications Mining on Automation Cloud, follow pre-requisites such as enabling the service, being an existing user, and obtaining an invite from the Admin.
🔒 Pre-requisites for accessing Communications Mining on Automation Cloud:
- Enable Communications Mining as a service on Automation Cloud tenant with an enterprise license and available AI units.
- Be an existing user on the Automation Cloud tenant; if not, request an Admin to add you.
📝 Process to obtain an account:
- An Admin on Communications Mining should create a new user profile for you through the ‘Manage Access’ page in the Admin console within the platform.
🚩 Tenant-based access and permissions:
- Access and permissions are specific to each tenant, not the entire organization. Ensure access is granted to the tenant where Communications Mining is enabled.
Note: The Admin cannot add you to Communications Mining until you accept your invite to the UiPath Automation Cloud.
Understand Permission in UiPath Communication Mining Account:
📝 Default user permissions for new users in a project:
- ‘View labels’ and ‘View sources’ permissions are granted by default.
- These permissions provide access to non-sensitive datasets, verbatims, and associated labels.
❌ Limitation of default permissions:
- Default permissions do not allow applying or removing labels within a dataset.
- Additional permissions, like ‘Modify users,’ are required, which can be granted by another user with such authority in the project.
💡 Admin access for Communications Mining via Automation Cloud:
- Admins on the cloud tenant automatically receive Admin access on Communications Mining.
- Admin privileges include ‘Sources permissions,’ ‘Datasets permissions,’ ‘Streams permissions,’ ‘Users permissions,’ ‘Buckets permissions,’ ‘Integrations permissions,’ and ‘Utility permissions.’
Navigating in UiPath Communication Mining Account:
🗺️ Platform Navigation:
- The platform has two main navigation menus: the navigation bar and the admin console in the top right-hand corner.
- The navigation bar becomes visible after selecting a dataset from the Datasets page.
- The admin console provides access to ‘Sources,’ ‘Integrations,’ ‘Manage Access,’ and ‘My Account.’
📚 Admin Console Elements:
- ‘Sources’ lead to the Sources page for creating or updating project sources.
- ‘Integrations’ allow creating and managing integrations.
- ‘Manage Access’ navigates to the Projects page for creating new projects or adding users to existing projects.
- ‘My Account’ leads to the My Accounts page for viewing account details and accessing API tokens.
📊 Dataset Navigation:
- After selecting a dataset, users with ‘View Sources’ permissions can access various pages using the top navigation bar, such as Train, Discover, Explore, Validation, Reports, Models, Streams, and Settings.
❓ Help Menu:
- The help menu, accessible via the ‘?’ symbol, provides access to various resources, including CM Support (knowledge base), CM API Docs (developer documentation), CM Training Academy (courses and training), UiPath community forum, UiPath Academy, YouTube tutorials, product release notes, product documentation and guides, and downloadable UiPath resources.
AI Units and Licensing Metering and Charging Logic:
💼 AI Units and Licensing:
- AI Units are used to license AI products and are charged based on consumption when models bring value.
- Consumption cost is calculated as the sum of prediction cost and hardware cost.
📈 Prediction Cost:
- Prediction cost is computed based on the input size and unit cost of the model.
- Different models have distinct unit costs and input size calculations.
💻 Hardware Cost:
- The hardware cost for deploying ML Skills is determined by replicas and resource costs.
- The resource cost varies depending on the CPU, RAM, and GPU.
🧾 Consumption Example:
Given the scenario, AI Units are calculated for training and hosting the model and making predictions using models.
for Example, – *Communication Mining Charges AI Unit based on “Number of Messages” or ” No Of Text in Tickes” (Per message uploaded, modified, or predicted)
See Information on Source Page – AI Center – AI Units (uipath.com)
How to Implement UiPath Communication Mining :
Let’s dive into the step-by-step process of its implementation:
🔍 Step 1: Data Collection and Preparation
The first step in implementing Communication Mining is collecting the relevant communication data. This data can include emails, chat logs, audio recordings, social media messages, and more. Ensure that you have permission to access and analyze this data in accordance with privacy and data protection regulations.
Here we need to connect our communication channel of unstructured data which can be sourced from the Support Mailbox, Workflow or ITSM Tickets, Survey Responses, and Feedback Forms.
Data into communication mining can be imported via:
- Live Integration using limited pre-built connectors such as Microsoft Exchange & Salesforce
- Using API Integration
- Or You can upload Historical Data exported from other systems via CSV or API
🎯 Step 2: Discover Data and Apply Labels
Once you have the required dataset the “UiPath Communication Mining Platform automatically starts the discovery process.
With the help of its unsupervised learning models, it clusters “Group of Communication” sharing similar themes. You can apply labels and entities that capture required data points.
This will act as the first stage of your Model training and provide a high-level view of “What’s in your Data …”
You should clearly define the objectives you want to achieve through the dataset.
🧠 Step 3: Train the Model
With the data loaded and objectives set, start utilizing the training modes to build out the training data for the model.
The platform enables any business user from the communication channel to become a Model Trainer without the requirement of data scientists or engineers.
With each training action, the platform continuously retrains, enhancing its comprehension of concepts and data points, leading to real-time updates in predictions.
By labelling a small, representative sample of training data, the platform gains the ability to automatically interpret and make predictions across the entire dataset at scale.
🧩 Step 4: Validate the Prediction
Structured label and entity predictions, each with their own confidence scores will be provided as part of the prediction. But before using this prediction in Automation Process or Take Decision we must validate the prediction.
Again the Tools functionality provides full transparency when it comes to performance, validating your model automatically each time it retrains.
With the help of “Model Rating” you will be able to make improvements as suggested as the next action item.
🤝 Step 5: Analyze, Monitor and Refine & Automate
The platform aggregates all of the predictions for labels and entities with key metadata to provide a wealth of query-able data, providing visibility into previously hidden processes and channels. Alongside analytics, we can deploy trained models to enable production automation.
Now Interpret the insights gained from Communication Mining and apply them to your business strategies and processes. Use the information to make data-driven decisions that align with your objectives.
UiPath robots and downstream systems can utilise the structured data created by UiPath Communications Mining to extend automation into service and conversation-based processes
As Communication Mining is an iterative process. Continuously monitor the results and refine your approach based on feedback and new data.
This ongoing refinement will ensure the tool’s effectiveness and relevance to your business needs.
Summary: The model retrain process in Communications Mining involves the creation of new model versions when labels or entity reviews are applied. The predictions are recomputed across the dataset, and the process can take time depending on training, dataset size, and the number of labels. Users can track the status of their model through helpful status indicators.
⚙️ Model Retraining: Applying labels or reviewing entities leads to the creation of new model versions.
♻️ Recomputation of Predictions: The model recomputes predictions across the dataset during retraining.
⏳ Time Required: The retraining process can take time, influenced by training progress, dataset size, and label count.
📊 Status Feature: Communications Mining provides a status feature to help users stay updated on their model’s progress.
✅ Up-to-Date Indication: The status icons at the top of the page indicate the current status of the dataset.
Build Your Automation Using Communication Mining Prediction :
So far we have covered the various steps required for using the Communication mining platform. Once you are happy with the Model training you can follow the below steps to utilize the model in your automation use case :
📌 Pinning the Model: The model needs to be pinned by toggling the “Save” option on the Models page.
⚙️ Stream Configuration: Specify the stream name, dataset name, model version, and labels of interest, along with their corresponding thresholds.
🔄 Fetch-and-Advance Loop: Use the stream to fetch comments from the platform, acknowledging previous batches with an advance request.
📝 Process Results: The response includes comments, predicted labels, and entities, which need to be parsed accordingly.
❗ Exception Handling: The exception endpoint allows tagging incorrectly predicted items as exceptions for model improvement.
NOTE: You can also use Get Predictions for a pinned model using Prediction API Route.
Read more – Predictions | Communications Mining Docs (reinfer.io)
FAQs (Frequently Asked Questions) :
The implementation timeframe depends on the size of your communication data and the complexity of your objectives. On average, the process can take a few weeks to a few months, considering data collection, preparation, and analysis.
Yes, UiPath Communication Mining supports multiple languages through its advanced NLP capabilities. It can analyze communication data in various languages, providing a global perspective on your business operations.
Currently, It supports all major European languages, but additional language support is coming soon.
Data privacy is a top priority when implementing Communication Mining. Ensure that you comply with all relevant data protection regulations and obtain necessary consent from individuals whose data is being analyzed.
Yes, UiPath Communication Mining can be integrated with your existing business intelligence tools, allowing seamless collaboration and data sharing across different departments.
Yes, UiPath Communication Mining is designed to be scalable, allowing you to analyze large volumes of communication data as your business grows.
