Human-Computer Dialogue Knowledge Fusion to Resolve Ambiguous Feedback

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing human-computer dialogue systems face issues with slow updates in text knowledge and knowledge graphs, leading to ambiguous or unclear feedback dialogue data due to missing key information and limited representation capabilities.

Innovation Solution

Integrate a combination of historical and current dialogue data to determine semantic keywords, utilizing a knowledge graph, text knowledge, and external knowledge sources such as network and multi-modal knowledge, with multi-hop retrieval and redundancy removal to enhance the reliability and validity of feedback dialogue data.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If text knowledge and knowledge graph are updated slowly, then the system structure remains stable, but the feedback dialogue data becomes ambiguous and unreliable

Engineering Contradiction:

Improvereliability of feedback dialogue dataVSAvoidupdate speed of knowledge base

Core Design Contradiction:

ReliabilityVSLoss of time

Solution Approach 1:

The knowledge base is segmented into multiple independent knowledge sources including knowledge graph, text knowledge, and external knowledge sources. This segmentation allows each knowledge source to be updated independently and queried separately, resolving the contradiction by enabling fast access to specific knowledge types without requiring complete base updates.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

A knowledge fusion module acts as an intermediary between multiple knowledge sources and the dialogue processing system. This intermediary integrates information from knowledge graph, text knowledge, and external sources, providing comprehensive and up-to-date information to improve feedback reliability without being constrained by slow updates of any single knowledge base.

Inventive Principle:

Principle #24Intermediary (Mediator)

2Reliability

If only knowledge graph and text knowledge are used, then the system complexity remains low, but the feedback dialogue data lacks comprehensiveness and becomes ambiguous

Engineering Contradiction:

Improvevalidity of feedback dialogue dataVSAvoidcomplexity of knowledge base structure

Core Design Contradiction:

ReliabilityVSDevice complexity

Solution Approach 1:

The system implements a universal knowledge query interface that can access multiple types of knowledge sources (knowledge graph, text knowledge, external knowledge) through a unified method. This multi-functional approach enhances feedback validity by drawing from diverse knowledge types while maintaining relatively simple system operation through standardized access patterns.

Inventive Principle:

Principle #6Universality (Multi-functionality)

Solution Approach 2:

The knowledge base system combines multiple types of knowledge sources (structured knowledge graph, unstructured text knowledge, and external knowledge) into a composite knowledge system. This composite structure improves feedback validity by leveraging the strengths of different knowledge types while the knowledge fusion module manages the complexity of integrating these diverse sources.

Inventive Principle:

Principle #40Composite materials

3Adaptability or versatility

If semantic keywords are determined without current dialogue environment data, then the processing speed remains fast, but the semantic keywords lack dimensionality and indication capability

Engineering Contradiction:

Improveindication capability of semantic keywordVSAvoiddialogue processing speed

Core Design Contradiction:

Adaptability or versatilityVSProductivity

Solution Approach 1:

The system performs preliminary determination of semantic keywords by combining historical dialogue data with current dialogue environment data before generating feedback. This preliminary action enriches the semantic keywords with multi-dimensional information from various sources, improving their indication capability while the optimized knowledge query process maintains efficient processing speed.

Inventive Principle:

Principle #10Preliminary action

Data Source

PatentUS20250252267A1Human-computer dialogue method, dialogue network model training method, and apparatus

Publication Date: 2025.08.07 HUAWEI TECH CO LTD

AI summary

The human-computer dialogue method includes: obtaining second historical dialogue data, current dialogue environment data, and second input dialogue data input by a user, determining a second semantic keyword based on the second historical dialogue data, the current dialogue environment data, and the second input dialogue data, and generating and outputting, based on the second semantic keyword and a preset-type knowledge base, second feedback dialogue data corresponding to the second input dialogue data, where the preset-type knowledge base includes: a knowledge graph, text knowledge, and external knowledge other than the knowledge graph and the text knowledge.