Skip to content

Taxonomy Builder Setup

This document outlines a process for creating, organizing, and managing topic hierarchies to categorize and analyze customer conversations effectively, aligning with business structures and analytical needs.

Prerequisites and Access Requirements

Before beginning taxonomy setup, ensure you have:

  • Full Conversation Intelligence Permissions: Only users with complete Conversation Intelligence access can configure taxonomies.

  • Advanced Topic Discovery Enabled: Taxonomy Builder requires the GenAI-powered topic discovery feature to be active in your application.

  • Business Structure Mapping: A clear understanding of your organizational hierarchy and customer interaction types.

Interface Architecture

Navigate the taxonomy builder interface effectively using the hierarchical tree view and organizational tools. Understanding the interface layout ensures efficient taxonomy creation and management workflows.

Navigation Path: Quality AI > Configure > Taxonomy Builder

Topics

Build and organize your conversation topics using a structured three-level hierarchy system. The topic interface provides visual tools to create, edit, and manage complex taxonomical relationships efficiently.

Understanding Topic Hierarchy Levels

The system supports three distinct hierarchical levels for topic organization:

  • L1 (Top Level): Parent categories representing broad topic areas.

  • L2 (Second Level): Child categories under L1 parents.

  • L3 (Third Level): Sub-child categories under L2 parents.

Key Feature: L2 nodes can be configured independently without requiring an L1 parent, providing flexibility in your taxonomy structure.
Topic Hierarchy Levels

Adding New Topic Level

  1. Click the Add New Topic Level button in the taxonomy interface.
    Add New Topic

  2. Select the appropriate hierarchy level (L1 or L2).
    Add New Topic Level

  3. Complete the required fields:

    • Name: Required, max 50 characters.

    • Description: Required, max 250 characters.

  4. Click Create to add the new level to your taxonomy.

Adding Second Level

  1. Click the + Add Level 2 button under the desired L1.
    Add Level 2

  2. Complete the required fields:

    • Name (required, max 50 characters)

    • Description (required, max 250 characters)

  3. Parent Category:

    • Set as Standalone Category (Level 2)

    • Assign to New Parent Category (Level 1)

  4. Click Create to add the intent to your taxonomy.

Add Third Level

  1. Click the + Add Level 3 button under the desired L2.
    Add Level 3

  2. Complete the required fields:

    • Name (required, max 50 characters)

    • Description (required, max 250 characters)

  3. Check the resolution detection option if you want to track successful/unsuccessful outcomes (enabled by default).

  4. Click Create to add the intent to your taxonomy.

Editing Existing Level

  1. Locate the level you want to modify in the taxonomy tree.

  2. Click the Edit option.
    Edit Existing Level

  3. Update the following fields in the modal window:

    • Name

    • Description

  4. Click Save to confirm your changes.

Resolution Configuration

Define success criteria and outcome measurements for your conversation topics and customer interactions. Configure resolution tracking to monitor the effectiveness of customer service interactions and identify opportunities for improvement.
Resolution Configuration

Setting Up Resolution Detection

Navigate to the Resolution Configuration tab to define success criteria for your intents.

Available Resolution Types:

  • Successful: Positive outcomes for the identified intent.

  • Unsuccessful: Negative or incomplete outcomes.

Customizing Resolution Definitions:

  1. Select the resolution type (Successful/Unsuccessful).

  2. Edit the default description to match your business requirements.

  3. Ensure descriptions account for:

    • Conversation type (sales/support)

    • Specific topics identified

    • Your organization's success criteria

      Note

      Default descriptions are provided based on conversation type and identified topics to streamline initial setup.

Version Management

Implement systematic version control to track taxonomy changes and maintain historical data integrity. Version management ensures seamless updates while preserving analytical continuity across different taxonomy iterations.

Saving and Managing Versions:

  1. After making changes to your taxonomy, save the version to impact runtime analytics.

  2. Access version history from the Topic Discovery dashboard.

  3. View the current version plus the three most recent previous versions.

  4. Previous versions display data only up to their respective update dates.

Organizing Your Taxonomy

Drag-and-Drop Functionality

  • Drag intents to new positions within the hierarchy.

  • Drop zones are highlighted when valid.

  • Invalid operations are clearly indicated.

  • Preview the new structure before confirming changes.

Level Changes: You can move intents between different levels (e.g., L2 to L1, L2 to L3) while maintaining their child relationships.

Important Considerations:

  • Moving a parent intent automatically moves all its children

  • The system validates against duplicate names in the destination location

  • Confirmation is required before completing any move operation

Use Case Example - Banking Sector

Implement a taxonomy structure for financial services using real-world banking scenarios and customer interaction patterns. This example demonstrates the practical application of hierarchical topic organization in a complex, regulated industry environment.

Phase 1: Business Structure Analysis and Planning

Organizational Mapping Exercise: Document your current business structure. For a regional bank, this strategic framework includes:

L1 Strategic Categories:

  • Retail Banking (individual customer services)

  • Commercial Banking (business customer services)

  • Mortgage Services (specialized lending division)

  • Investment Services (wealth management and advisory)

L2 Tactical Breakdown Under Retail Banking:

  • Checking Accounts (daily banking transactions and account management)

  • Savings Products (savings accounts, CDs, money market accounts)

  • Personal Loans (unsecured lending products)

  • Credit Cards (revolving credit products and services)

L3 Operational Details Under Credit Cards:

  • New Applications (application process, approval status, documentation)

  • Payment Issues (payment processing, due dates, payment methods)

  • Balance Transfers (transfer requests, promotions, processing)

  • Account Cancellations (closure requests, retention efforts, final processing)

  • Reward Inquiries (points balance, redemption, program questions)

Phase 2: Topic Creation and Configuration

Creating L1 Topics:

  1. Click Add Intent and select L1 level.

  2. Enter a clear, business-aligned name (for example, "Retail Banking").

  3. Write a description: "All customer interactions related to individual banking services, including personal accounts, loans, and credit products. Covers day-to-day banking needs for retail customers rather than business or commercial services."

Expanding to L2 Topics:

  1. Select the parent L1 topic when creating L2 entries.

  2. Use product-specific naming (e.g., "Credit Cards" rather than generic "Card Services").

  3. Include specific scope in descriptions: "Credit card-related inquiries, including new applications, existing account management, payment processing, rewards programs, and account modifications. Excludes debit card issues, which fall under checking accounts."

Detailed L3 Configuration:

  1. Select an appropriate L2 parent.

  2. Use action-oriented naming that reflects customer intent (for example, "Payment Issues" vs. "Payments").

  3. Create detailed, example-rich descriptions: "Customer experiencing problems making credit card payments, including failed online payments, payment processing delays, payment method changes, autopay setup issues. Look for phrases like 'payment won't go through,' 'card was declined for payment,' 'need to change payment method.'"

  4. Configure resolution tracking based on interaction type.

Phase 3: Resolution Configuration Strategy

Enable for Actionable Interactions:

  • Service Requests: Account changes, product applications, cancellations.

  • Problem Resolution: Technical issues, billing disputes, service restoration.

  • Transaction Support: Payment assistance, transfer issues, and account access problems.

Skip for Informational Interactions:

  • Balance Inquiries: Simple information requests with no follow-up needed.

  • Rate Information: Interest rate questions, fee schedules, and general product info.

  • Confirmation Calls: Appointment confirmations, payment confirmations, routine notifications.

Custom Resolution Definitions Example: For "Payment Issues" L3 topic:

  • Successful: "Customer completed payment, or payment issue was resolved during the call. Payment processing error was fixed, an alternative payment method was set up, or a technical issue preventing payment was corrected."

  • Unsuccessful: "Customer unable to complete payment or payment issue remains unresolved. Technical problems persist, payment method cannot be updated, or the issue requires escalation or a follow-up call."

Phase 4: Version Management and Deployment

Creating New Versions: All taxonomy modifications exist as drafts until you explicitly create a new version. This prevents accidental impacts to ongoing conversation analysis and gives you complete control over when changes take effect.

Version Deployment Process:

  1. Complete all desired taxonomy modifications.

  2. Review changes in the preview mode.

  3. Create a new version with a descriptive name and a change summary.

  4. The system automatically applies the new taxonomy to future conversation analysis.

  5. Historical data remains classified under previous taxonomy versions.

This technical implementation approach ensures your Taxonomy Builder deployment aligns with business objectives while maintaining the flexibility to evolve with organizational needs and customer interaction patterns.