LLM Query Conversion for Consistent Structured Search Filtering

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Conventional search systems face challenges in optimally selecting filters, leading to inefficient resource usage and inconsistent search results due to outdated filter definitions, evolving vocabulary, and limitations in user search flexibility, which can result in irrelevant or incomplete search outcomes.

Innovation Solution

A natural language processing (NLP) engine uses a large language model to convert natural language queries into structured search queries, automatically identifying filter types and values without user explicit selection, thereby enhancing search efficiency and consistency.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If conventional search systems use traditional filter selection methods, then users can apply filters to search results, but the system wastes computing resources retrieving irrelevant results and produces inconsistent search outcomes due to outdated filter definitions

Engineering Contradiction:

Improvesearch result consistencyVSAvoidcomputing resource waste

Core Design Contradiction:

ReliabilityVSLoss of energy

Solution Approach 1:

The patent replaces the mechanical filter selection system with an NLP-based semantic understanding system. Instead of relying on predefined filter definitions and manual filter application, the system uses natural language processing to interpret user intent and generate structured queries that automatically incorporate relevant filtering criteria, eliminating the need for traditional filter mechanisms

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The patent transforms the search system from using static filter definitions to dynamic semantic parameters extracted from natural language queries. The NLP engine analyzes user input to identify and extract relevant filtering parameters (e.g., location, time, category) directly from the query text, allowing the system to adapt to evolving vocabulary and user preferences without updating filter definitions

Inventive Principle:

Principle #35Parameter changes

2Adaptability or versatility

If search systems use predefined filter definitions, then filters can be applied to search results, but the system cannot adapt to evolving vocabulary and user search flexibility

Engineering Contradiction:

Improvevocabulary evolution adaptationVSAvoidfilter definition maintenance

Core Design Contradiction:

Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The patent makes the search system dynamic by replacing static filter definitions with dynamic natural language processing. The NLP engine continuously adapts to new vocabulary and user expressions by analyzing query patterns and extracting relevant parameters in real-time, allowing the system to evolve with user language without requiring updates to filter definitions

Inventive Principle:

Principle #15Dynamics

Solution Approach 2:

The system performs self-updating through the NLP engine, which automatically learns and adapts to new vocabulary and user search patterns. The NLP model continuously improves its parameter extraction capabilities by processing user queries, eliminating the need for manual filter definition maintenance while maintaining high adaptability

Inventive Principle:

Principle #25Self-service

3Ease of operation

If users manually select filters, then they can control search criteria, but the process is tedious and limits user search flexibility

Engineering Contradiction:

Improvesearch operation simplicityVSAvoiduser search flexibility

Core Design Contradiction:

Ease of operationVSAdaptability or versatility

Solution Approach 1:

The patent introduces an NLP engine as an intermediary between the user and the search system. The NLP engine translates natural language queries into structured search queries with appropriate filters automatically applied, serving as a mediator that understands user intent and bridges the gap between casual user input and precise search requirements, eliminating the need for manual filter selection while maintaining search control

Inventive Principle:

Principle #24Intermediary (Mediator)

Data Source

PatentUS12475150B2Configuring a large language model to convert natural language queries to structured queries

Publication Date: 2025.11.18 MICROSOFT TECHNOLOGY LICENSING LLC

AI summary

Embodiments of the disclosed technologies are capable of generating natural language queries. The embodiments describe generating a training natural language query of a training structured search query using a first LLM and a first prompt. The embodiments further describe fine-tuning a second LLM using the training natural language query of the training structured search query and the training structured search query. The fine-tuned second LLM generates a structured version of a natural language query. The embodiments further describe generating the structured version of a received natural language query using the fine-tuned second LLM and a second prompt.