OCR Link Detection and Correction for URL Safety
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing character recognition systems cannot effectively differentiate between URLs of risky websites and well-known websites, leading to potential conversion of image-based URLs into risky website URLs during optical character recognition (OCR), which may result in accessing unsafe content.
Innovation Solution
An information processing apparatus and method that recognizes characters in images, detects link information representing network addresses, corrects non-link character strings, and converts link information into character strings before correction, utilizing a blacklist and whitelist to verify the safety of accessed data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If OCR is applied to convert image-based URLs into electronic text, then character recognition capability is improved, but the risk of converting URLs into risky website URLs increases
Solution Approach 1:
The system performs preliminary detection of link information in the character string before correction is applied. By identifying URLs in the original character information and storing them for later reference, the system ensures that even if correction processes modify the text, the original URL structure can be recovered, preventing conversion to risky website URLs.
Solution Approach 2:
The system introduces an intermediary mechanism that stores the original link information separately from the corrected character string. This intermediary storage allows the system to reference the authentic URL structure after correction, acting as a mediator between the correction process and the final URL validation, thereby preventing harmful conversions.
2Measurement precision
If character correction is applied to improve OCR accuracy, then character recognition quality is improved, but link information may be incorrectly modified
Solution Approach 1:
The system performs preliminary detection and storage of link information before the correction process is applied to the character string. This ensures that the original URL structure is preserved in a separate storage location, allowing the correction process to improve character recognition quality without compromising the integrity of link information.
Solution Approach 2:
The system segments the character information into different components: link information (URLs) and non-link character strings. By applying correction processes selectively only to non-link portions while preserving link information, the system improves overall character recognition quality while maintaining the reliability and integrity of URL structures.
3Measurement precision
If all character strings are corrected to improve overall accuracy, then character recognition accuracy is improved, but URL structure may be distorted
Solution Approach 1:
The system segments character information into link information and non-link character strings, applying different processing strategies to each segment. Correction processes are applied selectively to non-link portions to improve overall character accuracy, while link information is preserved with its original structure intact, maintaining URL precision.
Solution Approach 2:
The system applies different quality standards and processing methods to different parts of the character information. Non-link character strings receive full correction processing to improve local accuracy, while link information maintains its original structural precision. This local differentiation ensures both overall accuracy improvement and URL structure preservation.
Data Source
AI summary
An information processing apparatus includes a hardware processor that: accepts an image input from outside as an input image; recognizes a plurality of characters in the input image and produces character information that is constituted by the characters and includes a plurality of character strings; detects, from the character information, link information representing a network address of data; and corrects one or more of the character strings other than a character string that constitutes the link information among the character strings.