Automated Answer Generation for Technical Support Queries
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current self-help and automated technical support systems for interconnected products are inefficient due to the burden on users to search for relevant information, as they require users to sift through numerous answers to find the correct solution.
Innovation Solution
A system that automatically provides solutions by extracting keywords from user questions, using a linguistic module to match them with a knowledge base and a machine learning model to predict answers based on previous data, with scores combining the frequency and probability of matches, and presenting solutions ordered by normalized scores.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If users browse or search knowledge bases and FAQs to find solutions, then users can access technical support information, but users must read through many answers and know what to search for, increasing user burden and time consumption
Solution Approach 1:
The system automatically generates answers by analyzing user questions against product documentation and code, eliminating the need for users to manually search through knowledge bases. The automated answer generation service performs the information retrieval task itself rather than requiring users to navigate and search
Solution Approach 2:
The patent replaces the mechanical search-and-browse process with an automated natural language processing system that uses machine learning models to generate answers directly from product documentation and code, substituting manual user actions with automated computational processes
2Adaptability or versatility
If comprehensive knowledge bases are provided for interconnected products, then more technical support information is available, but the system complexity and difficulty of finding relevant information increase
Solution Approach 1:
The patent introduces an automated answer generation service as an intermediary between the comprehensive product documentation and users. This service processes user questions, retrieves relevant information from the extensive documentation and code, and synthesizes answers, thereby managing the complexity of comprehensive technical support systems without exposing users to that complexity
Solution Approach 2:
The system extracts only the most relevant information from comprehensive product documentation and code repositories based on the specific user question, rather than presenting all available information. This extraction process filters out unnecessary complexity while maintaining comprehensive support coverage
3Ease of operation
If manual technical support is provided for connected products, then users receive personalized assistance, but the cost and time for providing support increases significantly
Solution Approach 1:
The system enables self-service technical support by automatically generating answers from product documentation and code, allowing users to receive personalized assistance without human intervention. The automated service analyzes individual user questions and generates tailored responses, eliminating the need for manual support while maintaining personalized service quality
Solution Approach 2:
The patent incorporates feedback mechanisms where user interactions with generated answers are used to continuously improve the automated answer generation service. This feedback loop enables the system to learn from real usage patterns and improve over time, maintaining high support quality while operating efficiently at scale
Data Source
AI summary
A system is disclosed for automatically providing solutions to user questions containing text and data. The system includes a linguistic module to extract keywords from the question text, lookup solutions from a knowledge base that match the keywords, with a score based on the keyword frequency and match, and use a machine learning model trained on previous question data to predict solutions from the question data, with score based on the model's probability, and a combination module for combining the solutions ordered by their respective normalized scores.