Knowledge graph retrieval and enhanced generation for customer service Q & A
It demonstrates the entire workflow of a system that uses knowledge maps to build, retrieve, and answer questions, especially in terms of technical support and problem solving. The following is a detailed description of each part:
knowledge map construction
Raw data: The input data in the figure are some technical support tickets.
Parsing and connection: The system first parses work order data and establishes connections between work orders. For example,”CLONE_FROM” means that one work order is cloned based on another work order.
Vector database: Work order information is converted into text, embedded and stored in the vector database for efficient similarity retrieval.
Search and question answering
Problem analysis: In practical applications, users may ask specific questions, such as how to solve specific technical problems.
Entity detection and intent classification: The system identifies key entities in the problem (such as “CSV upload error”) and user intentions (such as “Steps to Reproduce”).
Embedded search: Using the previously built vector database, the system retrieves the most relevant work orders related to the problem.
Answer generation: Based on the search results, the system generates steps or answers to solve the problem.
detailed steps
Internal work order tree analysis: Identify the internal structure of a single work order.
Work room connection: Establish the relationship between work orders.
Node value text embedding generation: Convert the text content of the work order into embedding vectors to support subsequent text similarity comparisons.
Entity detection and intent classification: Parses user queries, extracts key information, and classifies user query intentions.
Embedded retrieval and filtering: Use embedded vectors to retrieve relevant work orders and then filter them based on the specific needs of the issue, such as priority and issue type.
Answer generation: Generate the answers or action steps required by the user based on the retrieved information.
application scenarios
This type of system is often used in areas such as customer support and IT service management to automate common technical problems and improve the efficiency and accuracy of problem resolution. Through the construction of knowledge maps and efficient retrieval mechanisms, the system can understand and process complex user queries and provide accurate solutions.
If you want to learn more, you can click on the link below the video.
Thank you for watching this video. If you like it, please subscribe and like it. thank
Paper：https://arxiv.org/pdf/2404.17723
Video: