Online Clustering for Customer Support Topics
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing customer-support systems face challenges in automatically clustering and evolving customer-support topics over time, as they need to adapt to new tickets and changing issues without prior knowledge of the topics that will arise, and existing clustering methods are not suitable for real-time updates.
Innovation Solution
A system that uses online clustering with a sketch representation to assign new tickets to clusters based on distance thresholds, creating new clusters when necessary, and updates existing clusters incrementally, employing techniques like Word2vec, LSTM networks, and feed-forward neural networks to generate request vectors and maintain relevant self-service content.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If traditional clustering methods are used to group customer-support tickets, then tickets can be organized into topics, but the system cannot adapt to new topics in real-time and requires reprocessing all historical data when topics change
Solution Approach 1:
The patent implements dynamic clustering by maintaining multiple cluster representations (centroids, sketches, and boundary definitions) that can be incrementally updated as new tickets arrive. The system dynamically adjusts cluster boundaries and memberships without requiring complete reprocessing of historical data, allowing real-time adaptation to emerging topics while maintaining efficient organization of existing tickets.
2Measurement precision
If the system processes all tickets to form accurate clusters, then clustering precision improves, but the computational complexity and processing time increase significantly
Solution Approach 1:
The patent segments the clustering process into distinct components: vector generation from ticket text, distance calculation to cluster centroids, sketch-based boundary representation, and incremental cluster updates. Each component processes only the necessary subset of data (new tickets against existing cluster representations), avoiding the need to reprocess all historical tickets while maintaining clustering accuracy through the coordinated operation of these segmented functions.
3Loss of information
If the system maintains detailed information about all customer requests, then topic analysis becomes more accurate, but storage requirements and processing overhead increase
Solution Approach 1:
The patent extracts essential features from complete ticket texts by generating condensed vector representations that capture the semantic meaning of customer requests. Instead of storing and processing full text documents, the system extracts key information into fixed-dimensional vectors that can be efficiently stored, compared, and updated, significantly reducing storage requirements while preserving the information necessary for accurate topic clustering and analysis.
Data Source
AI summary
The system obtains a set of tickets representing customer requests for a customer-support ticketing system. Next, the system produces a set of request vectors by feeding words from each ticket through a model to generate a request vector for the ticket, wherein the set of request vectors is represented as a set of points in a vector space. The system then performs a clustering operation on the set of points to form clusters representing support topics, wherein the clustering operation creates a new point for a new ticket in the vector space when the new ticket is received, and incrementally updates existing clusters to accommodate the new point. Finally, the system presents a user interface to a customer-support agent, wherein the user interface uses the support topics to organize the customer requests, and enables the customer-support agent to perform a customer-support operation in response to a customer request.