Patent Figure Annotation via Visual Association
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing technologies face challenges in accurately recognizing and annotating patent figures due to complex font styles, colors, and layouts, leading to segmentation issues, recognition errors, and user confusion, as well as difficulties in navigating between figures and specifications.
Innovation Solution
A visual association process that uses multi-core processors to search and match references in patent figures with identifiers in the text, allowing for accurate recognition and annotation, and enabling users to visually associate identifiers with references through positioning, rotation, and color-based techniques for improved usability.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If OCR engines are used to recognize patent figures with various fonts and styles, then character recognition is attempted, but recognition accuracy deteriorates due to font complexity
Solution Approach 1:
The patent segments patent figures into multiple image regions based on visual characteristics, separating text annotations from technical drawings. This allows different processing approaches for different regions, improving OCR accuracy by focusing on text-specific areas while handling graphical elements separately.
Solution Approach 2:
The patent introduces an intermediary processing step that generates text annotations and metadata about figure contents, which then link to the main patent specification. This intermediary layer facilitates navigation and understanding without requiring direct OCR of all figure elements.
2Loss of information
If annotations are displayed over patent figures to provide context, then understanding is improved, but figure usability deteriorates due to visual obstruction
Solution Approach 1:
The patent moves annotations from the two-dimensional figure space to a third-dimensional hyperlink structure. Annotations are stored as separate metadata objects that can be accessed through interactive elements, allowing context information to be available without visually obstructing the figure.
Solution Approach 2:
The patent creates copies of figure references and annotations in multiple locations and formats. Hyperlinks and cross-references are placed strategically throughout the patent document, providing access to annotation information without requiring permanent overlays on the original figures.
3Loss of information
If patent figures contain many part numbers to provide detailed information, then completeness is improved, but understanding deteriorates due to complexity
Solution Approach 1:
The patent implements a universal linking system where part numbers in figures are connected to corresponding text descriptions, claims, and specifications through hyperlinks. This multi-functional approach allows users to access detailed information about any part number through a single interface, reducing the cognitive load of tracking multiple numbers.
Solution Approach 2:
The patent provides feedback mechanisms through hyperlinked annotations that respond to user interactions. When users click on part numbers or figure references, the system provides immediate feedback by displaying related text, specifications, or cross-references, helping users understand the significance of each element.
4Measurement precision
If patent specification is written in technically dense language to provide precise information, then accuracy is improved, but readability deteriorates
Solution Approach 1:
The patent performs preliminary organization of technical information by structuring the specification with clear headings, paragraphs, and cross-references before presentation to users. Hyperlinked annotations are pre-configured to guide users through the dense technical content in a logical sequence, making complex information more accessible.
Data Source
AI summary
Various computing technologies for content analysis.