Ellice Group said on the 29th that it has released an AI document analysis solution, Helpy Document Vision, which lets artificial intelligence (AI) analyze complex forms of documents on its own and convert them into structured data.
Helpy Document Vision can automatically analyze not only paragraphs within documents but also a variety of visual elements such as tables, charts, formulas, and images. The company said it secured high accuracy and throughput based on Helpy Table Vision, a vision-language model (VLM) with strengths in data analysis.
According to Ellice Group, Helpy Table Vision can accurately analyze old documents or modern and contemporary history documents that general models find hard to recognize, using training data on the order of only a few hundred cases. It also offers features that provide precise structured data, including analysis of long Excel documents with hundreds of rows, restoration of chart and graph data (HTML), and encoding of complex formulas.
While existing solutions took an average of 33.6 seconds for document layout analysis and data extraction, Helpy Document Vision recorded an average of 9.8 seconds, showing a processing speed about 3.4 times faster, the company said.
An Ellice Group official said, "Helpy Document Vision organically combines in-house data and training infrastructure, enabling rapid optimization tailored to domain-specific documents in finance, healthcare, and law," adding, "Through this, it can convert the vast unstructured data within corporations into high-quality digital data and improve the performance of retrieval-augmented generation (RAG) and AI agent systems that corporations have recently been adopting."
Ellice Group plans to advance the VLM technology used this time beyond document recognition into a vision-language-action model (VLA) capable of situational judgment and action execution. VLA is a technology that helps AI precisely control robots or machines on actual industrial sites by reading complex manuals or drawings, and is expected to play a key role in "physical AI," where AI operates in the physical environment.
Kim Su-in, chief risk officer (CRO) of Ellice Group, said, "Through Ellice's AI document analysis solution, we will help corporations experience real workflow automation innovation by converting complex documents that were handled manually into high-quality data."