LLM Query Reformulation with Ontology Filtering

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Natural language processing systems face challenges in processing user prompts and understanding user intent due to ambiguity, context, and language variations, while efficiently managing and accessing data from various sources, including ontologies and databases, is also complex.

Innovation Solution

A computing architecture is implemented that uses a multi-stage natural language processing pipeline, integrating a generative AI model to reformulate ambiguous queries and an ontology-driven approach to provide structured domain knowledge, along with few-shot learning techniques to enhance response accuracy.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If a large language model is used to process natural language queries, then the system can understand user intent and generate responses, but the model encounters challenges with ambiguous queries leading to incorrect interpretations

Engineering Contradiction:

Improvenatural language understandingVSAvoidquery interpretation accuracy

Core Design Contradiction:

Adaptability or versatilityVSReliability

Solution Approach 1:

The patent introduces an intermediary processing layer between the user's ambiguous natural language query and the large language model. This layer includes query rewriting mechanisms and context enrichment components that transform the original ambiguous query into a more precise form before it reaches the LLM, thereby improving interpretation accuracy without sacrificing natural language understanding capabilities

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The system performs preliminary actions by pre-processing the natural language query through multiple stages including ambiguity detection, query rewriting, and context retrieval before the main LLM processing. This preliminary preparation reduces the ambiguity that the LLM would otherwise have to handle, leading to more accurate responses

Inventive Principle:

Principle #10Preliminary action

2Reliability

If an ontology-driven approach is used to provide structured domain knowledge, then the system can improve response accuracy, but computing resource utilization increases

Engineering Contradiction:

Improveresponse accuracyVSAvoidcomputing resource utilization

Core Design Contradiction:

ReliabilityVSUse of energy by moving object

Solution Approach 1:

The patent applies local quality by selectively retrieving and applying only the relevant portions of the ontology that pertain to the specific query context, rather than loading or processing the entire ontology. This targeted approach maintains response accuracy by using appropriate domain knowledge while minimizing computing resource utilization by avoiding unnecessary processing of irrelevant ontology elements

Inventive Principle:

Principle #3Local quality

Solution Approach 2:

The ontology is segmented into modular, query-relevant portions that can be independently retrieved and processed. This segmentation allows the system to load only the necessary subset of ontology knowledge required for each specific query, reducing overall computing resource requirements while maintaining the structured domain knowledge needed for accurate responses

Inventive Principle:

Principle #1Segmentation

3Ease of operation

If the chatbot interface is decoupled from the reporting interface, then the user experience is enhanced with real-time updates, but the system complexity increases

Engineering Contradiction:

Improveuser experienceVSAvoidsystem architecture

Core Design Contradiction:

Ease of operationVSDevice complexity

Solution Approach 1:

The system is segmented into distinct functional modules: a chatbot interface for natural language interaction, a reporting interface for data visualization, and a backend processing layer that coordinates between them. This modular segmentation enables real-time updates and enhanced user experience while managing system complexity through clear separation of concerns and defined interface contracts between components

Inventive Principle:

Principle #1Segmentation

Data Source

PatentUS20250139151A1Data structure modification using a virtual assistant based on large language models and ontology

Publication Date: 2025.05.01 ADP INC

AI summary

A system can receive, via a chatbot interface, a textual input related to an electronic report. The system can generate, using a large language model, an output including a set of keywords, a rephrased version of the textual input, and an intent sentence. The system can filter, using the output from the large language model, an ontology stored in a database to generate a filtered ontology, including one or more ontology elements, intents, and examples. The system can generate, using the large language model and the filtered ontology, a plurality of actions that are compatible with the electronic report. The system can display, via the chatbot interface, the actions. The system can receive, via the chatbot interface, an indication to execute an action. The system can provide, responsive to the indication, instructions to execute the action on the electronic report to modify the electronic report.