Page Stream Segmentation via Structural Pattern Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current page stream segmentation tools require prior knowledge of document contents and are ineffective for documents without page numbers or barcodes, and struggle with new document formats or foreign languages, often necessitating user input and re-programming for accurate boundary determination.
Innovation Solution
A system that performs language-agnostic page stream segmentation by detecting characters on each page, computing structural data based on position and font format, and using machine learning to determine page boundaries without user input, allowing automatic segmentation of multi-page files into documents.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If existing tools use pre-programmed rules with static indicators (page numbers, barcodes, bookmarks), then page boundary determination can be performed, but the tools require prior knowledge of document contents and cannot handle new document formats or foreign languages
Solution Approach 1:
The patent transforms the approach from using fixed content-based indicators (page numbers, specific text strings) to analyzing structural parameters of document pages. By computing structural data such as character positions, font formats, text density, and layout patterns, the system adapts to different document formats and languages without requiring prior knowledge of their specific contents. This parameter-based analysis enables the tool to reliably identify page boundaries across diverse document types.
Solution Approach 2:
The patent replaces the mechanical rule-based system (checking for specific indicators like barcodes or predefined text strings) with an intelligent system that uses machine learning models. These models analyze structural patterns and characteristics of pages to automatically determine boundaries, eliminating the need for pre-programmed rules and enabling adaptation to new document formats and languages without reprogramming.
2Reliability
If tools require user input of prior knowledge about document contents, then accurate segmentation can be achieved for known formats, but the ease of operation decreases and user expertise is required
Solution Approach 1:
The patent implements a self-service system where the page stream segmentation tool automatically analyzes document structural data and performs segmentation without requiring user input about document contents. The machine learning model independently evaluates structural characteristics of each page and determines boundaries, making the tool easy to operate while maintaining high accuracy across different document types without requiring user expertise.
3Ease of manufacture
If static pre-programmed rules are used for page boundary determination, then implementation is simple, but the system cannot accurately determine boundaries for new document formats or languages
Solution Approach 1:
The patent creates a universal page boundary detection system that handles multiple document formats and languages through a single machine learning model. Instead of implementing separate rules for each document type, the system uses structural pattern recognition that works across diverse formats, making the implementation broadly applicable while maintaining adaptability to new formats without requiring separate programming for each case.
Data Source
AI summary
Described are system, method, and computer-program product embodiments for performing language-agnostic page stream segmentation. In some embodiments, a method includes receiving a multi-page file associated with a plurality of documents. A plurality of characters present on each page of the set of consecutive pages, including a first page and a second page, of the multi-page file can be detected. A plurality of structural data for each page can be computed based on a position and a font format for one or more of the detected characters. The plurality of structural data between the first page and the second page can be compared to determine whether the second page corresponds to a boundary between two documents of the plurality of documents. The multi-page file can be segmented at the second page in response to determining that the second page corresponds to the boundary.