Retrieval-Augmented Conversation Flow Routing for Accurate Answers

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

State-of-the-art machine learning based language models (LLMs) provide generic and often misleading responses due to hallucinations, failing to meet the specific needs of certain domains where accurate answers are critical, and lack context-specific information.

Innovation Solution

An online system manages conversations using a machine learning based language model that routes interactions through predefined conversation flow types, utilizes organization-specific data sources, and performs critical analysis to generate context-specific and accurate responses.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If generic training data is used to train language models, then the model can provide answers applicable to a wide context, but the answers are not helpful when questions need to be answered within a specific context

Engineering Contradiction:

Improveapplicability to wide contextVSAvoidcontext-specific accuracy

Core Design Contradiction:

Adaptability or versatilityVSReliability

Solution Approach 1:

The system segments the knowledge base into multiple data sources with different levels of generality. It retrieves and prioritizes specific context-related information over generic training data by querying external knowledge bases and ranking results based on relevance to the user's context, thereby resolving the contradiction between wide applicability and context-specific accuracy

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The system introduces an intermediary retrieval-augmented generation layer between the user query and the language model. This intermediary component fetches context-specific information from external data sources and injects it into the model's context window, enabling the model to answer specific context questions accurately while maintaining its general knowledge capabilities

Inventive Principle:

Principle #24Intermediary (Mediator)

2Productivity

If language models are trained on large amount of generic data, then they can provide generic answers, but they suffer from hallucination and provide misleading information

Engineering Contradiction:

Improveresponse generation capabilityVSAvoidinformation accuracy

Core Design Contradiction:

ProductivityVSReliability

Solution Approach 1:

The system implements feedback mechanisms where retrieved context information is used to verify and correct the language model's generated responses. The system checks whether the model's answers are consistent with the retrieved factual information and can prompt the model to revise hallucinated content, thereby maintaining productivity while improving reliability

Inventive Principle:

Principle #23Feedback

Solution Approach 2:

The system performs preliminary retrieval of context-specific information from external data sources before the language model generates its response. By having the correct information available in advance and providing it as context to the model, the system prevents hallucination from occurring in the first place, maintaining both response generation capability and information accuracy

Inventive Principle:

Principle #10Preliminary action

3Reliability

If organization-specific data is integrated into the conversation system, then context-specific and accurate responses can be provided, but the system complexity increases

Engineering Contradiction:

Improveresponse accuracyVSAvoidsystem architecture complexity

Core Design Contradiction:

ReliabilityVSDevice complexity

Solution Approach 1:

The system uses a universal retrieval-augmented generation architecture that can handle multiple types of data sources (internal knowledge bases, external APIs, documentation, etc.) through a single unified interface. This multi-functional design allows the system to integrate organization-specific data without proportionally increasing complexity, as the same core mechanism handles diverse data types

Inventive Principle:

Principle #6Universality (Multi-functionality)

Solution Approach 2:

The system introduces an intermediary layer that abstracts the complexity of organizing-specific data integration. This intermediary component handles data retrieval, validation, and formatting, shielding the core conversation logic from data source complexity while enabling accurate context-specific responses through structured information injection

Inventive Principle:

Principle #24Intermediary (Mediator)

Data Source

PatentUS20250335718A1Online conversation flows guided by machine learning based language models and retrieval augmented generation

Publication Date: 2025.10.30 WISQ INC

AI summary

A system performs routing of conversation flow routing for an online conversation. The online system stores metadata describing a plurality of conversation flow types. Each conversation flow type comprises a sequence of steps describing natural language-based interactions with a user. The system generates a prompt comprising a natural language request and metadata describing conversation flow types and requests a machine learning based language model to identify a particular conversation flow type relevant to the natural language request. The system provides the prompt to the machine learning based language model for execution and receives a response identifying a conversation flow type relevant to the natural language request. For subsequent natural language requests, the system follows the steps of the identified conversation flow type and generates a reply based on steps of the identified conversation flow type.