Ink Document Parsing with Stroke- and Object-Level Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Parsing digital ink documents with complex layouts, including overlapping and diverse content items, is challenging due to varying user writing styles and computing resource constraints, particularly on handheld devices.
Innovation Solution
A hybrid approach combining stroke-level and object-level processing using neural networks for pixel classification and bounding box detection, with shared computational resources, to accurately parse ink documents.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If stroke-level processing is performed on all ink strokes to classify each individual stroke, then parsing accuracy is improved, but computational time and processing resources increase
Solution Approach 1:
The patent segments the ink document into distinct objects first, then applies stroke-level processing only to strokes within each object. This segmentation approach allows the system to process strokes in smaller, manageable units rather than analyzing every stroke independently across the entire document, thereby reducing overall computational time while maintaining parsing accuracy.
Solution Approach 2:
The patent applies stroke-level processing selectively only to strokes that are needed for accurate parsing, rather than uniformly processing all strokes. By identifying which strokes are critical for distinguishing between objects and which can be processed at a higher level, the system achieves accurate parsing with reduced computational effort.
2Productivity
If object-level processing is performed to group ink strokes into objects, then parsing efficiency is improved, but the ability to handle complex layouts with overlapping objects deteriorates
Solution Approach 1:
The patent segments the processing into two distinct levels: object-level processing for grouping strokes into candidate objects, and stroke-level processing for precise classification. This segmentation allows the system to first establish high-level object structures efficiently, then refine the classification of individual strokes within those structures, thereby maintaining both efficiency and accuracy in complex layouts.
Solution Approach 2:
The patent transitions from a single-level processing approach to a two-dimensional processing architecture where object-level and stroke-level analyses operate in different dimensions. The object-level processing handles the macro structure and grouping, while stroke-level processing addresses the micro details and classification. This dimensional separation enables the system to efficiently handle complex layouts by addressing different aspects of the problem simultaneously at appropriate levels of detail.
3Measurement precision
If neural networks are used for pixel classification and bounding box detection, then parsing accuracy is improved, but computing resource requirements increase
Solution Approach 1:
The patent segments the computational workload by applying neural networks only at the stroke-level processing stage for classification, rather than using them throughout the entire parsing pipeline. The object-level processing uses less computationally intensive methods for grouping strokes, thereby reducing overall resource consumption while maintaining accuracy through targeted neural network application where it is most needed.
Solution Approach 2:
The patent applies computationally intensive neural network-based pixel classification and bounding box detection only to the extent necessary for accurate stroke classification within objects, rather than uniformly applying these heavy computations across the entire document processing pipeline. This selective application reduces total computing resource consumption while preserving parsing accuracy in the critical areas where neural networks provide the most value.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Technology is descried herein for parsing an ink document having a plurality of ink strokes. The technology performs stroke-level processing on the plurality of ink strokes to produce stroke-level information, the stroke-level information identifying at least one characteristic associated with each ink stroke. The technology also performs object-level processing on individual objects within the ink document to produce object-level information, the object-level information identifying one or more groupings of ink strokes in the ink document. The technology then parses the ink document into constituent parts based on the stroke-level information and the object-level information. In some implementations, the technology converts the ink stroke data into an ink image. The stroke-level processing and/or the object-level processing may operate on the ink image using one or more neural networks. More specifically the stroke-level processing can classify pixels in the input image, while the object-level processing can identify bounding boxes containing possible objects.