Book a Meeting

Technology: Retrieval Augmented Generation (RAG)

InsightGPT is an AI based software based on the Retrieval Augmented Generation framework.

flow-insightgpt

The RAG framework is an innovative approach in natural language processing (NLP) that combines the strengths of retrieval-based and generation-based models. In RAG, the system first retrieves relevant information from a large database (vector store) using a retrieval mechanism.

This retrieved information is then used to augment the generation process, where the model generates responses or outputs based on both the retrieved information and the input query or context.

This framework enhances the contextual understanding of the model by leveraging existing knowledge from the retrieved data, leading to more accurate and coherent responses. RAG is particularly effective in tasks where access to relevant context is crucial for generating high-quality outputs.

The quality of the entire process, which encompasses both the retrieval and generation phases, is monitored through the measurement of a set of key performance indicators.

  • Context Precision: conveys the quality of the retrieval pipeline, assessing the relevancy of the retrieved context to the prompt.
  • Context Recall: measures the ability to retrieve all necessary information comprehensive.
  • Faithfulness: evaluates the extent of hallucinations in the generated answers, ensuring factual consistency against the given context.
  • Answer Relevancy: measures how ‘to the point’ answers are to the question, assessing the relevance of the generated answer to the prompt.
  • Answer Correctness: evaluates the accuracy and reliability of the responses generated by a system or model. It encompasses two main aspects.
  • Semantic Similarity: measures the degree of similarity between the generated answer and the expected response in terms of meaning, understanding, and context.
  • Factual Similarity: evaluates the accuracy of the information provided in the generated answer against factual knowledge or ground truth. It assesses whether the information presented is factually correct and aligned with the context of the question or prompt.

Book a Meeting 📅

My Agile Privacy
✕
This website uses technical and profiling cookies. Clicking on "Accept" authorises all profiling cookies. Clicking on "Refuse" or the X will refuse all profiling cookies. By clicking on "Customise" you can select which profiling cookies to activate.
AcceptRefuseCustomise

Learn how we helped 100 top brands gain success