Automated Help Center Article Suggestion via Vector Clustering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing customer-service systems face challenges in determining which articles need to be written to resolve ongoing customer-service issues, especially in new systems without existing help center articles, and in addressing commonly occurring or new issues not adequately covered by existing articles.
Innovation Solution
A system that automatically suggests articles to be written by generating request vectors from customer tickets, embedding them in a vector space, performing clustering operations, and notifying content creators when new or updated articles are needed to cover identified clusters, using machine-learning techniques such as recurrent neural networks and Word2vec to analyze and prioritize clusters based on distance and relevance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If help center articles are manually created to resolve customer-service issues, then customer-service quality improves, but the time and resources required to identify and write articles increases
Solution Approach 1:
The system enables automatic identification of article needs through self-service mechanisms. The machine learning model autonomously analyzes customer-service tickets, identifies gaps in existing help center articles, and generates suggestions for new articles without requiring manual intervention from support staff.
Solution Approach 2:
The patent replaces the manual mechanical process of reviewing tickets and deciding what articles to write with an automated machine learning system. The ML model processes tickets, performs clustering analysis, and generates article suggestions automatically, substituting human effort with computational processes.
2Ease of operation
If comprehensive help center articles are created to cover all possible issues, then customer self-service effectiveness improves, but the complexity and cost of maintaining the help center increases
Solution Approach 1:
The system performs preliminary analysis to identify only the specific articles that are needed based on actual customer-service tickets. Rather than creating comprehensive coverage of all possible issues, the ML model proactively identifies gaps in existing articles and suggests only the necessary new articles, reducing unnecessary complexity.
Solution Approach 2:
The patent dynamically adjusts the help center content based on changing customer needs. The machine learning model continuously analyzes incoming tickets and adjusts article recommendations accordingly, allowing the help center to adapt its scope and depth based on actual demand rather than maintaining fixed comprehensive coverage.
3Measurement precision
If human agents review and prioritize article creation requests, then article relevance improves, but the processing speed and scalability decrease
Solution Approach 1:
The system incorporates feedback mechanisms where the machine learning model learns from the analysis of customer-service tickets and the outcomes of article creation. The model uses feedback from ticket patterns, clustering results, and article performance to continuously improve its relevance assessments while maintaining high processing speed.
Solution Approach 2:
The patent replaces human agent review with automated machine learning processes for prioritizing article creation requests. The ML model autonomously assesses relevance based on ticket analysis and clustering, eliminating the bottleneck of manual review while maintaining or improving relevance through data-driven insights.
Data Source
AI summary
The system obtains a set of tickets representing customer requests generated by a customer-support ticketing system. Next, the system feeds words from each ticket through a model to generate a request vector for the ticket, wherein the request vector comprises numerical values representing words in the ticket. The system then embeds the request vectors in a vector space. If help center articles already exist, the system embeds article vectors for the existing help center articles in the vector space. Next, the system identifies clusters of request vectors, which are within a pre-specified distance of each other in the vector space. If an identified cluster is more than a pre-specified distance away from a closest article vector in the vector space, the system notifies a content creator that a new article needs to be written, or an existing article needs to be updated, to cover the identified cluster.