Ledger Sheet Recognition Using Reference Point Character Strings
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing ledger sheet recognition techniques face challenges in accurately extracting multiple character strings associated with a single character string, particularly in complex layouts, leading to difficulties in correctly identifying and extracting relevant information.
Innovation Solution
An information processing apparatus and method that identifies a reference point character string based on relative positions and uses it to extract corresponding character strings, improving the accuracy of ledger sheet recognition by associating item names with their respective values.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional character recognition is used to extract character strings from ledger sheets, then basic text extraction is achieved, but accuracy deteriorates when one character string is associated with multiple character strings in complex layouts
Solution Approach 1:
The patent segments the character string extraction process into multiple stages: first extracting all character strings, then identifying reference point character strings, and finally extracting corresponding character strings based on relative positions. This segmentation allows the system to handle complex layouts by processing information in manageable steps rather than attempting to extract all associations simultaneously.
Solution Approach 2:
The patent introduces reference point character strings as intermediary elements that mediate between item names and their corresponding values. By using reference points as intermediate markers, the system can accurately associate character strings even in complex layouts where direct association is difficult to determine.
2Reliability
If simple character extraction is used, then processing speed is maintained, but the ability to correctly identify and extract multiple corresponding character strings deteriorates
Solution Approach 1:
The patent performs preliminary extraction of all character strings from the ledger sheet image before performing the association process. This preliminary action creates a complete set of extracted text elements that can then be reliably associated through subsequent reference point matching, ensuring no character strings are missed in the final extraction.
Solution Approach 2:
The patent transitions from simple linear text extraction to two-dimensional spatial relationship analysis by using relative positions of character strings on the ledger sheet. By incorporating spatial dimensionality (horizontal and vertical positions), the system can reliably identify associations between item names and their corresponding values even when they appear in different locations on the page.
Data Source
AI summary
An information processing apparatus includes circuitry. The circuitry receives an input of ledger sheet image data read from a ledger sheet. The circuitry identifies, among a plurality of character strings extracted through character recognition performed on the ledger sheet image data, a reference point character string based on relative positions of the reference point character string and a particular character string of the plurality of character strings. The reference point character string forms a reference point for extraction. The circuitry extracts a corresponding character string corresponding to the reference point character string based on relative positions of the corresponding character string and the reference point character string.