Document Item Extraction via Visual Rendering and Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing document processing technologies face difficulties in accurately extracting and separating individual items from documents with non-conforming and non-standard formats, as the underlying structure can vary significantly while appearing visually similar, making it challenging to identify patterns and separate items from other data.
Innovation Solution
A system and method that utilizes a network-based provider with client-server architecture to render documents, break them into subdocuments based on visual elements, and append metadata for indexing and retrieval, allowing for efficient extraction and storage of item data from documents like web pages.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional document processing methods are used to extract data from non-standard documents, then the processing approach is simple, but the extraction accuracy is poor due to varying underlying structures
Solution Approach 1:
The patent uses rendering to create a visual copy of the document that exposes the actual layout and structure. By comparing the rendered visual elements with the underlying code structure, the system can identify patterns and accurately extract items even from non-standard documents. This copying approach allows the system to bypass the complexity of varying document formats while maintaining high extraction accuracy.
Solution Approach 2:
The system changes the representation parameters of the document by rendering it into visual elements with spatial coordinates, bounding boxes, and hierarchical relationships. This parameter transformation converts unstructured or semi-structured document data into a standardized format that can be systematically processed and extracted, resolving the accuracy issue without requiring complex document-specific processing logic.
2Adaptability or versatility
If multiple items and data are included on a single document page for viewer convenience, then the document is more useful to users, but the difficulty of separating individual items and their data increases
Solution Approach 1:
The patent segments the document into individual items and their associated data by analyzing the rendered visual elements and their spatial relationships. The system identifies boundaries between different items using bounding boxes and hierarchical structures from the rendering process, then separates them into distinct extractable units. This segmentation enables accurate extraction of individual items while preserving the multi-item layout that provides user convenience.
3Adaptability or versatility
If the underlying code structure of a document varies significantly while maintaining visual similarity, then the document format is flexible, but pattern identification and accurate data separation become difficult
Solution Approach 1:
The patent creates a rendered copy of the document that visualizes the actual layout regardless of the underlying code structure. By comparing this visual representation with the code, the system can identify the true structural patterns and relationships. This approach maintains flexibility in accepting various document formats while achieving accurate pattern identification through visual verification.
Solution Approach 2:
The system adds a visual dimension to the document analysis process by rendering the document and extracting spatial information (coordinates, bounding boxes, layout positions). This dimensional transformation allows the system to identify patterns based on visual structure rather than relying solely on code structure, enabling accurate pattern identification even when code formats vary significantly.
Data Source
AI summary
A system receives a document including a plurality of items. The system then processes the document to expose a plurality of item elements associated with one or more items of the document, the item elements including visual information used to render the document. The system then identifies each item of the one or more items based on one or more of the plurality of item elements and corresponding visual information.