Virtual Agent Training Using Ranked Optimal Utterances

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

The current training of virtual agents relies heavily on manually identified phrases, which is inefficient and lacks automation.

Innovation Solution

A method and system for training virtual agents by storing conversations in logs, mining utterances, calculating scores based on syntactic and semantic similarity, and ranking them to extract optimal utterances for training.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If manual identification of training phrases is used, then training data can be obtained, but the process is inefficient and time-consuming

Engineering Contradiction:

Improvetraining efficiencyVSAvoidtime for phrase identification

Core Design Contradiction:

ProductivityVSLoss of time

Solution Approach 1:

The system enables automated self-training by having the virtual agent independently identify, score, and select its own training phrases from conversation logs without human intervention. The automated scoring mechanism using syntactic and semantic similarity allows the system to self-evaluate and self-improve, replacing manual phrase identification entirely.

Inventive Principle:

Principle #25Self-service

Solution Approach 2:

The patent replaces the manual mechanical process of phrase identification with an automated computational system. The scoring mechanism using algorithms for syntactic similarity (comparing sentence structures) and semantic similarity (comparing meanings) substitutes human analysts, dramatically improving efficiency while maintaining or enhancing training quality.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

2Extent of automation

If automated utterance mining is implemented, then training phrase identification is automated, but system complexity increases

Engineering Contradiction:

Improveautomation of phrase identificationVSAvoidcomplexity of scoring system

Core Design Contradiction:

Extent of automationVSDevice complexity

Solution Approach 1:

The automated scoring system is divided into distinct modular components: syntactic similarity scoring (analyzing sentence structure) and semantic similarity scoring (analyzing meaning). This segmentation allows each component to be developed, tested, and maintained independently, managing overall system complexity while achieving high automation.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The scoring system serves multiple functions: it evaluates utterance quality, ranks candidates for training, and adapts to different conversation contexts. By creating a universal scoring framework that handles various utterance types and scenarios, the system achieves high automation without proportionally increasing complexity.

Inventive Principle:

Principle #6Universality (Multi-functionality)

Data Source

PatentUS12541650B2Method and system for training a virtual agent using optimal utterances

Publication Date: 2026.02.03 QUANTIPHI INC

AI summary

A method and system for training a virtual agent is provided herein. The method and system comprises storing conversations between the virtual agent and a user in logs. The method and system further comprises mining the logs to retrieve utterances. The method and system further comprises computing regression for each of the plurality of the charging time segments. The method and system further comprises providing a score to the utterances. Further, the method ranking the utterances based on the score.