Attending ECAI – the European Conference on Artificial Intelligence held in Spain – the research team from Viettel Artificial Intelligence and Data Service Center (Viettel AI) introduced a solution that enables real-time, automated extraction of table structures. This innovation accelerates data extraction up to four times faster than current domestic and international solutions on the market.

Automated data extraction leverages technologies like artificial intelligence (AI) and optical character recognition (OCR) to automatically capture information from sources such as text, images, or scanned documents and convert it into easily manageable formats, like Excel files. This function is essential in digital office applications, playing a critical role in the digitization of physical documents for organizations. Although text recognition has been relatively successful, accurately identifying and extracting information from tables in documents remains challenging. Automating this process reduces manual data entry, increasing accuracy and processing speed.

According to a representative from the research team, Viettel AI’s table data extraction solution processes information at speeds up to 40 FPS (frames per second) in some cases, quadrupling the speed of current technologies. Importantly, while enhancing speed, accuracy remains comparable to existing solutions, with only about a 2% variance on standard datasets. Unlike traditional two-step extraction processes, the team streamlined this into a single-step approach, allowing faster handling of tables with multiple rows and columns. The solution also simplifies the processing complexity, optimizing memory use and making AI model training more efficient, paving the way for further advancements.

The research team representative also shared that this technology is now integrated into Viettel IDP – Viettel AI's smart document processing solution. Viettel IDP can automatically extract information from images at speeds below 2 seconds per page, 60 to 80 times faster than manual data entry, with up to 90% accuracy, helping users save 80% of the time in document approvals. The technology unveiled at ECAI 2024 marks the first step in accelerating Viettel IDP’s processing speed, aiming for not just 2-second but instant information processing.

ECAI is one of the world's most reputable recurring AI conferences, attracting hundreds of distinguished experts, researchers, and scientists from multiple countries. The event showcases the latest research and technologies, acting as a launchpad for breakthrough AI solutions and ideas. ECAI 2024 is co-organized by the European Association for Artificial Intelligence and the Spanish Association for Artificial Intelligence.

Viettel AI, a subsidiary of Viettel Group, leads in developing AI, Big Data, Robotics, and Digital Twin products and services. Viettel AI’s ecosystem comprises many top-quality products in Vietnam, trusted by numerous major organizations and enterprises domestically and internationally.