Customer Service Ticket Similarity Using Feedback-Updated Encoding Model

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing customer service ticket systems struggle to efficiently identify and manage duplicate or related tickets, leading to increased workload and delayed resolution times for support teams.

Innovation Solution

A method utilizing self-supervised learning to generate encodings of customer service tickets in a vector space, determining pairwise similarities, obtaining user feedback, updating similarities, and generating an updated encoding model using supervised learning to improve ticket similarity determination.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If traditional ticketing systems are used to manage customer service requests, then all tickets can be recorded and tracked, but duplicate and related tickets cannot be efficiently identified, leading to increased workload and delayed resolution

Engineering Contradiction:

Improveticket resolution efficiencyVSAvoidresolution time

Core Design Contradiction:

ProductivityVSLoss of time

Solution Approach 1:

The patent replaces manual ticket review and comparison processes with an automated encoding model that transforms ticket descriptions into vector representations. This mechanical substitution enables automatic similarity detection between tickets, allowing the system to identify duplicates and related issues without human intervention, thereby improving resolution efficiency and reducing time loss

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The patent transforms ticket data from unstructured text to structured vector encodings in a continuous vector space. By changing the representation parameters of ticket information and applying similarity metrics, the system can efficiently compute relationships between tickets, enabling rapid identification of duplicates and related issues that would be impossible to detect through traditional text-based methods

Inventive Principle:

Principle #35Parameter changes

2Measurement precision

If manual review of tickets is performed to identify duplicates, then accurate identification can be achieved, but the process becomes time-consuming and labor-intensive

Engineering Contradiction:

Improvesimilarity determination accuracyVSAvoidreview time

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The patent substitutes manual ticket review with an automated encoding model that computes vector representations and similarity metrics. This mechanical substitution maintains measurement precision by using sophisticated embedding techniques and similarity calculations, while eliminating the time-consuming nature of manual review processes

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The system performs self-service by automatically encoding tickets, computing similarities, and identifying duplicates without requiring human reviewers. The encoding model autonomously processes ticket data and generates similarity assessments, freeing support staff from manual review tasks while maintaining accurate identification capabilities

Inventive Principle:

Principle #25Self-service

3Ease of manufacture

If self-supervised learning is used to generate initial encodings, then the model can be trained without labeled data, but the encoding accuracy for similarity determination is insufficient

Engineering Contradiction:

Improvemodel training easeVSAvoidencoding accuracy

Core Design Contradiction:

Ease of manufactureVSMeasurement precision

Solution Approach 1:

The patent applies preliminary action by first training the encoding model using self-supervised learning to generate initial encodings without requiring labeled data. This preliminary training phase establishes a baseline model that can process tickets, which is then refined through supervised fine-tuning using labeled similarity data to achieve the required encoding accuracy for production use

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The system maintains continuity of useful action by transitioning from self-supervised pre-training to supervised fine-tuning without interruption. The model continuously learns and improves its encoding capabilities, first from unlabeled data to establish basic functionality, then from labeled data to refine accuracy, ensuring uninterrupted model development and deployment

Inventive Principle:

Principle #20Continuity of useful action

Data Source

PatentUS12314878B2Customer service ticket similarity determination using updated encoding model based on similarity feedback from user

Publication Date: 2025.05.27 DELL PROD LP

AI summary

Techniques are provided for customer service ticket similarity determination using an updated encoding model based on similarity feedback from a user. One method comprises obtaining encodings of customer service tickets in a vector space using an encoding model; determining pairwise similarities for the encodings of the customer service tickets; obtaining feedback from a user regarding the pairwise similarities for a subset of the encodings; updating pairwise similarities for the subset of the encodings using the feedback from the user; generating an updated encoding model by processing the updated pairwise similarities for the subset of the encodings of the customer service tickets using a supervised learning algorithm; and processing at least one customer service ticket based at least in part on the updated encoding model. The feedback from the user may indicate a similarity of two or more of the plurality of customer service tickets.