Preeti
The report discusses a system that combines Retrieval-Augmented Generation (RAG) with a Knowledge Graph (KG) to enhance customer service chatbots. Traditional chatbots rely on keyword matching, which can be ineffective for complex questions. The proposed system uses a KG to build a network of interconnected information from past customer service tickets and RAG to retrieve relevant information and generate responses. This system improves retrieval, enhances answers, and overall service effectiveness.
The system automatically answers customer service questions by organizing knowledge from past tickets and understanding questions using vector similarity. It leverages cosine similarity between embedding vectors for implicit connections. The Neo4j GDS library provides measures to calculate similarity between arrays. The system generates embeddings for graph node values using pre-trained text-embedding models like BERT and E5.
The research aims to enhance automated question answering systems by developing an automated mechanism for extracting graph templates, enhancing system adaptability, updating the knowledge graph based on user queries, and exploring the system’s applicability in other contexts beyond customer service. The system combines two techniques to find answers more effectively, improving overall customer service. The researchers also seek to make the system adaptable, continuously learning, and cost-effective for resource-limited scenarios.
Source link
Source link: https://mobinshaterian.medium.com/rag-system-knowledge-graph-the-system-for-finding-the-best-answer-in-the-tickets-management-e2ae2692c7c6?source=rss——llm-5
GIPHY App Key not set. Please check settings