Document Image Information Extraction System Based on Optical Character Recognition

Abstract

A document image information extraction system based on optical character recognition includes a storage unit and a processing unit. The storage unit stores a document image and a data form related to the document image. The data form includes a plurality of data fields and first information filled in at least a portion of the data fields, the first information originating from the document image. The processing unit is electrically connected to the storage unit to access the storage unit. The processing unit reads the document image and the data form from the storage unit, performs optical character recognition on the document image to generate text information, performs text cleaning and date and number formatting on the text information to generate pre-processed text information, and analyzes the pre-processed text information to extract a plurality of second information related to the data fields included in the data form from the pre-processed text information. The processing unit then corrects and completes the data in the data fields of the data form containing the first information based on the second information.

Claims

1. A document image information extraction system based on optical character recognition, comprising: A storage unit stores a file image and a data form associated with the file image, the data form including a plurality of data fields and first information filled in at least a portion of the data fields, the first information originating from the file image; and a processing unit electrically connected to the storage unit for accessing the storage unit; the processing unit includes an optical character recognition module, a data preprocessing module, a large language model, and a data postprocessing module; wherein the processing unit reads the file image and the data form from the storage unit, and executes the optical character recognition module to perform optical character recognition on the file image to generate text information; the processing unit executes the data preprocessing module to perform text cleaning and date and number formatting on the text information to generate preprocessed text information; The processing unit executes the large language model, causing the large language model to analyze the pre-processed text information to extract a plurality of second information related to the data fields contained in the data form from the pre-processed text information; the processing unit executes the data post-processing module, causing the data post-processing module to compare the second information with the first information contained in the data form, so as to correct and complete the data containing the first information in the data fields of the data form according to the second information.

2. The document image information extraction system based on optical character recognition as described in claim 1, wherein, The text cleaning action performed by the data preprocessing module on the text information includes removing useless characters from the text information; the date and number formatting action performed by the data preprocessing module on the text information includes converting the date and number in the text information from full shape to half shape and padding the date.

3. The document image information extraction system based on optical character recognition as described in claim 1, wherein, The data preprocessing module also replaces time pronouns in the text information, adds keyword tags to the text information, and adds dates to the text information if it determines that there is no date in the text information.

4. The document image information extraction system based on optical character recognition as described in claim 1, wherein, The large language model extracts the second information from the pre-processed text information based on pre-defined prompts.

5. The document image information extraction system based on optical character recognition as described in claim 1, wherein, When the first and second pieces of information are related to doctor's orders, the data post-processing module corrects and completes the data in the data fields of the data form by removing duplicate and invalid records from the data fields, cleaning and merging duplicate record names from the data fields, forcing a specific treatment recorded in the data form to be recorded on a single day, adding an outpatient date to the data form when it is determined that the data form has a recorded surgery date but no recorded outpatient, emergency, or inpatient date, completing the recorded inpatient or emergency date in the data form, converting the data form into Chinese fields and Gregorian calendar to generate a formatted data form, and storing the formatted data form in the storage unit.

6. The document image information extraction system based on optical character recognition as described in claim 1, wherein, The document image information extraction system based on optical character recognition is a computer device, and the optical character recognition module, the data preprocessing module, the large language model and the data postprocessing module are contained in a software program stored in the storage unit and can be read and executed by the processing unit.

Images