Intent-Based Deep Search with LLM Query Expansion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Search engines often return a multitude of irrelevant or less useful results, with more relevant information buried or not included at all, necessitating improved search methodologies.
Innovation Solution
Implementing deep search functionality using generative artificial intelligence (AI) models like large language models (LLMs) to generate intents and alternative queries, assign relevance scores to search results, and sort them based on these scores for enhanced relevance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional search utilities are used to retrieve information, then search speed is maintained, but relevance and usefulness of results deteriorate
Solution Approach 1:
The patent introduces LLMs as intermediary components between the user query and the search utility. The LLMs generate multiple alternative queries that capture different intents and aspects of the original query, which then query the search utility. This intermediary layer transforms a single query into multiple refined queries, improving result relevance without sacrificing information completeness.
Solution Approach 2:
The patent segments the single search query into multiple alternative queries by using LLMs to identify different intents and perspectives. Instead of relying on one broad search query that returns mixed results, the system breaks down the query into several targeted queries (e.g., breaking down 'best smartphones' into queries about camera quality, battery life, processing speed, etc.), each addressing a specific aspect of user need.
2Measurement precision
If multiple alternative queries are generated using LLMs, then search result relevance is improved, but system complexity increases
Solution Approach 1:
The patent employs a single LLM component that performs multiple functions: generating alternative queries, scoring search results for relevance, and ranking results. This multi-functional approach reduces the need for separate specialized components for each task, thereby managing system complexity while achieving improved search relevance through multiple query generations and intelligent result ranking.
Solution Approach 2:
The system implements feedback loops where LLMs evaluate and score search results based on relevance to the original query and alternative queries. The scored results are then ranked and fed back to the user in an optimized order. This feedback mechanism allows the system to automatically refine and improve result presentation without requiring complex manual intervention or additional hardware components.
3Measurement precision
If LLMs are used to generate relevance scores for each search result, then result accuracy is improved, but processing time increases
Solution Approach 1:
The patent applies partial action by having LLMs score and rank only the top portion of search results rather than processing every single result in exhaustive detail. The system generates alternative queries, retrieves results, and then uses LLMs to evaluate and rank the most relevant subset of results first, providing accurate ranking for the most important results while avoiding excessive processing time that would result from evaluating every single result with equal depth.
Data Source
AI summary
Systems and methods are provided for implementing deep search functionality using large language models (“LLMs”). In various examples, a computing system uses at least one LLM to generate intents based on a user query, to generate alternative queries based on a selected or identified primary intent, and to generate a relevance score for each search result that is obtained from a search utility (e.g., an Internet search engine, a file storage search utility, an email search utility, or a document storage search utility) in response to a primary query (corresponding to the primary intent) and the generated alternative queries being entered into the search utility. The search results from the search utility are sorted based on the corresponding generated relevance scores, and the sorted search results are caused to be displayed to the user as a deep search response to the user query.