Retrieval Augmented Generation (RAG)
Retrieval capabilities to enhance generative AI
How does it work?
RAG is used as a component within a broader and more complex system, typically a search engine or a conversational agent.
It consists of two modules that collaborate and interact with each other to provide accurate responses to user queries.
Modules
Retrieval module
When the system receives a user query, the retrieval module searches the knowledge base for the most relevant documents. This ensures better quality and reliability of responses compared to systems based solely on LLMs.
Generation module
The generation module formulates a coherent and appropriate response to the query using the information retrieved by the other module.
The use of an LLM makes the entire system more fluent and natural, both in understanding and generating text.
Application examples
Customer support
Provides accurate and relevant answers to customer inquiries
Educational tools
Enhances the efficiency of learning platforms, facilitating the learning process
Virtual assistants
Generates more precise outputs, improving chatbot response effectiveness
Content creation
Assists in content development by generating relevant text based on a given prompt
Healthcare support
Aids decision-making processes by retrieving and generating relevant information from medical literature
Our approach
In the AIWave platform, RAG is built on an architecture that integrates additional technologies to enhance natural language processing capabilities.
Information retrieval functionalities are more effective thanks to the use of ontologies and domain-specific dictionaries.
This architecture enables the retrieval of information from external sources, further reducing the workload on generative models, conserving computational resources, and ensuring greater reliability and control over responses.
