Miso Information Technology, a specialist in multimodal data platforms, said on Monday it will launch ViiX, a third-generation optical character recognition solution based on a vision-language model.

The company said ViiX is a domain-specific OCR solution that uses an LLM to understand and convert into data various documents generated in industrial settings, including hospital medical records and documents in manufacturing and construction.

Based on technology that analyzes document layout, fonts, spatial structure and context, it goes beyond simple text recognition to structure documents into data and provides an AI-based document-processing environment that links to search, analysis and task automation.

First-generation OCR focused on character recognition, while second-generation OCR supported deep learning-based table-area recognition and field extraction. But it required retraining whenever new document formats were added and had limits as costs rose to handle exception cases.

The company said ViiX, a third-generation OCR, improves key-value extraction accuracy by analyzing document context and structure together.

Miso Information Technology CEO Sang-do Nam (남상도) said, "ViiX is the result of systematizing with AI technology the domain knowledge Miso Information Technology has accumulated over the past 20 years in industrial settings." He added, "We will go beyond simply reading documents to understand the business context contained in them and usher in an era of document AI that implements task automation in the field."