The Korea Electronics Technology Institute (KETI) said on the 19th that it has completed development of the Next-generation AI Safety Report System, a technology that automatically selects, classifies, and forwards important reports submitted to the Ministry of the Interior and Safety's Safety Report Portal.
The technology was pursued as part of the government's national agenda to establish a disaster safety management system to ensure public safety. The government has emphasized the introduction of AI (artificial intelligence) for efficient processing as reports submitted to the Safety Report Portal exceed 10 million a year and continue to surge annually.
In response, the KETI research team joined with LG AI Research, URACLE, and TTA to develop the technology. They established a classification system for report types specialized for the safety report domain so AI can understand and categorize various reports on its own. They also built a high-quality training dataset and used a domestically developed AI foundation model so that it automatically generates and classifies report texts using only submitted photos.
In particular, KETI built a safety knowledge ontology based on its in-house data processing and unstructured data classification capabilities. An ontology is a network of relationships among concepts that AI can understand.
In addition, KETI applies this to a technology that filters report types of policy significance and immediately trains on corrections made by practitioners who directly participate in the classification process. This enables continuous improvement of classification performance even when the classification system is incomplete.
The Next-generation AI Safety Report System will begin field verification with the goal of launching a pilot service at the end of this year. In the verification phase, a next-generation AI reporting application will be developed to link the photo-based report text automatic generation model, and a report management AI agent that supports advanced forwarding classification will be developed and gradually applied to the Safety Report Portal system.
The KETI research team expects the Next-generation AI Safety Report System to reduce administrative burden and operating expense and to speed up report processing. They also said that once automatically classified report data accumulate, analysis by time, region, and frequency will be possible, which is expected to expand the establishment of preventive safety policies.
Moon Jae-won, principal researcher and project lead at KETI's AI Data and Security Research Center, said, "We closely combined the expertise of industry, academia, and research centered on KETI's AI and data capabilities," adding, "We will successfully introduce the AI Safety Report Portal specified in the national agenda and expand the verified technology into the private industrial sector."