Generative AI in Knowledge Discovery

Knowledge Discovery has a wide range of techniques for matching and returning information from an index of documents. In addition, you can use Large Language Models (LLMs) to enhance this process further and generate answers to queries that contain entirely new text, formed from analyzing important pieces of information from the index. This process is usually known as Generative AI, and Knowledge Discovery has a number of capabilities that use generative AI technology.

Vector Search

MainTopic: Vector Search

You can integrate vector search into the Content component, and run vector queries alongside or instead of conceptual search.

A vector is an array of numbers that represents an object, which might be text or an image. You can generate vectors by using many different models, and index them into Content in configured VectorType fields.

RAG Search

Main Topic: RAG Search

RAG search (Retrieval Augmented Generation) uses an LLM to generate answers from a set of data that it uses as context. RAG allows you to search within an index or set of documents, and then use an LLM to generate an answer from this context.

In Knowledge Discovery, you perform RAG search by using Natural Language Question Answering, which you configure with a RAG system. When a user sends a question, it retrieves documents from your Knowledge Discovery index that contain possible answers, and then sends these results as context to an LLM that you configure, in a prompt. The LLM then generates the answer from the context documents.

Search Abstractor

Many Knowledge Discovery generative AI options are included in a solution called Search Abstractor, which helps to streamline the process of handling natural language commands of different types. For example, it can determine when a user is asking a question that requires a natural language answer from your index, and triggers a RAG search. Alternatively, it can determine when the user is asking to retrieve particular documents, and runs a normal query.

Image Description

Image description in Media Server uses an AI model to generate a textual description of an image or video frame. For more information, refer to the Media Server Help.