Generative search

As explained in AI-Powered Search

Generative search produces search responses or results with a generative model instead of only retrieving existing documents.

Generative search creates responses from search results instead of returning only existing documents. It extends traditional ranked links and extractive answers toward synthesized summaries and answers. section 1.3.5 section 15.2

A typical flow retrieves and ranks relevant content, then uses a large language model to summarize multiple results or answer a question by synthesizing them. Retrieval-augmented generation supplies current context to reduce unsupported answers. section 1.3.5 section 1.4

Generative search commonly supports:

• Results summarization, which rewrites several ranked results into a concise response and can include citations. section 1.3.5 section 15.2
• Abstractive question answering, which generates an answer by synthesizing information from one or more search results. section 1.3.5 section 15.2
• Conversational and multimodal search experiences built with foundation models. section 15.0

Planning a personalized trip

For “What is an optimal agenda for three days in Paris assuming I really like fine dining, historic buildings and museums, but hate crowds?”, generative search can synthesize a tailored response from dynamically analyzed results. That flexibility also creates risks when generated responses reflect political or commercial bias, censorship, propaganda, or inaccurate synthesis. section 15.2

Figure 15.2 Spectrum of traditional search approaches (left) to generative search (right)

Use the spectrum diagram to compare traditional ranked links, predetermined answers, extractive question answering, results summarization, and abstractive question answering. section 15.2

How AI-Powered Search covers it

How deep it goes

Distinguishes extractive answers from generated synthesis but stops short of detailed implementation, prompt optimization, and model evaluation.

Examples

Uses ranked-result summaries, synthesized answers, citations, and a personalized three-day Paris itinerary query.

What it assumes

Builds from traditional information retrieval and extractive question answering while introducing synthesis through language models.

What you can do after

Evaluate how search can progress from retrieved documents and extracted answers to synthesized responses, including associated bias risks.

How it connects

Connects retrieval-augmented generation, large language models, extractive question answering, semantic search, and results summarization.

liveBookmanning.com

Read AI-Powered Search on liveBook Get the book at manning.com

Related concepts

• language modeluses · A language model is a computational model that represents patterns in linguistic tokens to estimate sequence probabilities and predict or generate language from context.
• result summarizationincludes · Result summarization is the process of condensing one or more search results into a shorter, more readable account while preserving their information.
• similarity searchimplements · Similarity search is a retrieval technique that finds or ranks items by comparing their representations with a query according to a similarity or distance measure.
• retrieval-augmented generationused by · Retrieval-augmented generation is a technique that retrieves relevant information from external sources and supplies it as context to a generative model when producing an output.
• abstractive question answeringproduces · Abstractive question answering synthesizes information into newly generated answer text rather than selecting text directly from source documents.

This explainer was generated by Manning's knowledge graph from the text of AI-Powered Search. The section links point to the parts of the book it draws on.