Flow Diagram Recognition Using Semantic Node-Edge Estimation

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional machine learning models struggle with accurately recognizing flow diagrams due to unclear start and end points of edges and intersections, leading to misrecognition and erroneous inference.

Innovation Solution

A computer program that utilizes a dialogue system with a machine learning model to estimate first and second information items, including nodes and edges, by inputting directives for semantic estimation and intersection recognition, improving accuracy through chain-of-thought reasoning.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If a conventional machine learning model is used to recognize flow diagrams, then the recognition process is simple and fast, but the accuracy of recognizing edge connections decreases when start and end points are unclear or edges intersect

Engineering Contradiction:
Improveedge connection recognition accuracyVSAvoidrecognition system complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The recognition process is divided into multiple sequential stages: node recognition first, then edge recognition. Each stage focuses on specific information items (nodes or edges) to improve accuracy. This segmentation allows the system to handle complex flow diagrams by breaking down the recognition task into manageable parts with directed attention.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The system performs preliminary node recognition and semantic estimation before edge recognition. By first identifying and semantically analyzing nodes, the system prepares contextual information that guides subsequent edge connection recognition, improving accuracy when edges are ambiguous or intersecting.

Inventive Principle:
Principle #10Preliminary action

2Measurement precision

If semantic estimation is performed for all information items simultaneously, then comprehensive recognition is achieved, but accuracy decreases due to lack of focused attention on difficult-to-recognize elements

Engineering Contradiction:
Improveinformation item recognition accuracyVSAvoidrecognition processing time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The system segments the recognition process into distinct phases: node information recognition, semantic estimation, and edge information recognition. Each phase directs computational resources to specific information items, improving accuracy without requiring simultaneous processing of all elements.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The system employs periodic action by repeatedly refining recognition results through multiple passes. The dialogue system iteratively processes information items, allowing time for accurate recognition while maintaining focus on difficult elements through repeated semantic estimation and validation.

Inventive Principle:
Principle #19Periodic action

3Measurement precision

If edge recognition is performed without considering node semantic information, then the process is faster, but misrecognition occurs when edges are unclear or intersect

Engineering Contradiction:
Improveedge connection recognition accuracyVSAvoidrecognition process complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The system merges node recognition results with edge recognition by integrating semantic information from nodes into the edge connection analysis. This combination allows the system to use contextual semantic cues from nodes to disambiguate edge connections, improving accuracy without requiring completely separate processing systems.

Inventive Principle:
Principle #5Merging (Combining)

Solution Approach 2:

Semantic information acts as an intermediary between node recognition and edge recognition. The dialogue system uses semantic estimates of nodes to guide and constrain edge connection recognition, providing contextual guidance that improves edge accuracy without directly observing ambiguous edge features.

Inventive Principle:
Principle #24Intermediary (Mediator)

Data Source

PatentEP4703964A1Computer program, information processing method, and information processing apparatus
Publication Date: 2026.03.04 FUJITSU LTD
  • EP4703964A1 patent drawingFigure 1
  • EP4703964A1 patent drawingFigure 2
  • EP4703964A1 patent drawingFigure 3

AI summary

An information processing apparatus inputs, to a dialogue system using a machine learning model, a first directive instructing estimation of a plurality of first information items based on data including the plurality of first information items and a plurality of second information items representing the relationships between the plurality of first information items. The information processing apparatus inputs, to the dialogue system, a second directive instructing semantic estimation of the first information items based on third information related to the meaning of the first information items. The information processing apparatus inputs, to the dialogue system, a third directive instructing estimation of the second information items based on the data. The information processing apparatus acquires, from the dialogue system, output information generated based on the second information items estimated in response to the third directive.