Handwriting Recognition Confidence Thresholding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current handwriting input conversion methods to machine typeset have low accuracy due to poor handwriting quality, resulting in false positives and false negatives, and lack feedback on recognition confidence, making it difficult for users to identify searchable content.
Innovation Solution
A method that determines the recognition confidence level for each handwriting object and only converts those with a recognition confidence level above a predetermined threshold into machine typeset, allowing users to adjust the threshold using recognition policies or a slider mechanism.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If handwriting input is converted to machine typeset without confidence threshold filtering, then conversion completeness is improved, but accuracy deteriorates due to false positives and false negatives
Solution Approach 1:
The patent applies parameter changes by introducing a confidence threshold parameter that filters handwriting objects based on their recognition confidence levels. By adjusting this threshold parameter, the system can control the balance between conversion completeness and accuracy, converting only those handwriting objects that meet the confidence criterion while excluding ambiguous cases that would cause false positives or negatives
Solution Approach 2:
The patent implements feedback by providing users with confidence level information for each converted handwriting object. This feedback mechanism allows users to understand the reliability of each conversion and make informed decisions about whether to accept or review specific conversions, thereby improving overall system reliability while maintaining productivity
2Reliability
If confidence threshold is increased to improve accuracy, then false positives and negatives are reduced, but conversion completeness deteriorates
Solution Approach 1:
The patent applies dynamics by making the confidence threshold adjustable rather than fixed. Users can dynamically change the threshold value based on their specific needs - increasing it when accuracy is prioritized or decreasing it when completeness is more important. This dynamic adjustment capability allows the system to adapt to different operational requirements without sacrificing either accuracy or completeness permanently
3Speed
If all handwriting objects are converted without confidence assessment, then conversion speed is improved, but quality deteriorates
Solution Approach 1:
The patent applies preliminary action by performing confidence assessment on each handwriting object before committing to the conversion. This preliminary evaluation step ensures that only handwriting objects with sufficient confidence are converted, preventing low-quality conversions from occurring in the first place. The confidence check is performed upfront, maintaining conversion quality without requiring extensive post-processing review
Data Source
AI summary
One embodiment provides a method, including: receiving, at an information handling device, an indication to convert handwriting input to machine typeset, wherein the handwriting input comprises one or more handwriting objects; determining, using a processor, a recognition confidence level for each of the one or more handwriting objects; and converting, response to the determining, each of the one or more handwriting objects having a recognition confidence level above a predetermined confidence threshold to one or more corresponding machine typeset words. Other aspects are described and claimed.