NetSmile Secures Patent for Advanced Document Image Processing Using Generative AI

In a significant advancement for the field of document processing, NetSmile, Inc., under the leadership of President Fukumitsu Saito, has successfully obtained a patent (Patent No. 7738872) for a sophisticated document image processing system that leverages generative AI technology. This patent, granted on September 5, 2025, is a testament to the potential of AI-driven solutions in transforming traditional OCR methods.

Overview of the Patent

The patent covers a groundbreaking technology that enhances the functionality of the company’s generative AI-OCR service, known as "なんでも読めるくん®" (Nandemo Yomerukun). This innovative approach resolves many of the challenges associated with conventional OCR systems, such as accurately digitizing various complex document formats.

Background of the Innovation

As the use of generative AI (particularly large language models) in the AI-OCR sector has grown, challenges such as "hallucination"—where incorrect information is produced—have surfaced. This has hindered the extraction accuracy of various document images, especially those with unstructured formats like invoices and estimates. NetSmile's newly patented technology addresses this problem, significantly improving the reliability of text recognition.

Key Features of the Invention

The newly acquired patent integrates traditional OCR methods with generative AI, allowing for mutual verification and correction of each other’s outputs to achieve high-precision text recognition. Here are the standout features:

1. Dual OCR Processing and Verification: The local OCR system scans the document images and generates text data simultaneously while the generative AI-OCR sends the same images and prompts to its server. It then compares results from both sources to determine accuracy; generated AI outputs are only accepted if deemed correct.

2. Feedback Mechanism for Accuracy Improvement: In instances of determined errors, the local OCR result is used to query the generative AI again with additional prompts. This process provides the generative AI with more information, thus enhancing the accuracy of its outcomes.

3. Location Information Identification: The local OCR also identifies the coordinates of the recognized text, correlating the generative AI outcomes with their positions on the document.

4. Use of Multiple Generative AI Models: While processing, several generative AI models work simultaneously to produce results, compensating for the weaknesses of any one model and yielding more reliable outputs overall.

With this patented technology, NetSmile has successfully tackled the issue of inconsistent extraction accuracy present in generative AI-OCR, paving the way for greater automation and efficiency in document processing tasks. Applications anticipated include invoice processing, contract management, and data conversion for various application forms.

About the Generative AI-OCR "なんでも読めるくん®"

Launched in April 2025, "なんでも読めるくん®" is the next-generation OCR service developed by NetSmile that utilizes generative AI for improved document processing capabilities.

Key Features:

- No Templates Required: Users can simply register the item names they wish to extract, allowing for automatic extraction from various formats of documents.
- Support for Diverse Formats: This service can effectively read and interpret various types of documents, including faded handwritten text and those containing complex layouts or symbols.
- Understanding and Summarizing: Beyond basic character recognition, it can summarize and generate information based on contextual understanding.

Benefits of Implementation:

- Significant reduction in processing time for invoices, contracts, and application forms.
- Alleviates the burden of template creation and management.
- Enhanced reading accuracy minimizes the need for manual checks.
- Anticipated implementation in diverse industries including manufacturing, logistics, retail, and finance.

With the incorporation of this patent technology, "なんでも読めるくん®" can deliver even more precise and reliable document data conversion, further accelerating businesses’ journey towards digital transformation.

About NetSmile

- Company Name: NetSmile, Inc.
- Location: 4-1-11 Yushima, Bunkyo-ku, Tokyo, Japan
- Representative: President Fukumitsu Saito
- Established: October 2013
- Capital: 100 million JPY
- Website: NetSmile
- Business Content: Generative AI-OCR, DX solutions, business automation using generative AI, machine learning.