[Edaily Reporter Lee So-Hyun ] Even in shipyards where the constant noise of welding and equipment operations is accompanied by a mix of regional dialects and on-site slang, artificial intelligence (AI) is providing real-time interpretation and translation support. FLITTO Inc.(300080)has partnered with Hanwha Ocean and HANWHA SYSTEMS to develop an AI interpretation and translation solution tailored to shipbuilding sites, which is being applied to safety training and work instructions for foreign workers.

FLITTO Inc. announced on the 16th that it has developed “OTalk,” a customized AI interpretation and translation solution for Hanwha Ocean, and implemented it at shipyard sites.

OTalk is an AI translation and interpretation solution planned by Hanwha Ocean’s AX Business Division based on on-site requirements and jointly developed by HANWHA SYSTEMS and FLITTO Inc. It goes beyond simply applying general translation services by incorporating technical terminology, slang, dialects, and even the noisy environment actually encountered in shipyards.

Hanwha Ocean collected data on shipbuilding-specific terminology, slang, and dialects used on the job site and compiled requirements based on the actual work environment and job characteristics. Based on this, FLITTO Inc. built a customized glossary and refined the AI translation model.

The system was also trained on voice data in seven major languages, including Korean, English, Vietnamese, and Thai. In particular, it underwent separate training to recognize the Gyeongsang Province accent and on-site pronunciation, which were difficult to understand accurately in the early stages. The company explained that it improved accuracy to the point where it can even recognize instances where “tank” is pronounced as “dangk.”

Sound sensitivity adjusted to match workshop noise levels

Features tailored to the unique noise environment of shipyards were also implemented. Based on feedback from on-site users and test results, Hanwha Ocean identified necessary features and, in collaboration with HANWHA SYSTEMS and FLITTO Inc., developed a “5-level sound sensitivity adjustment” function tailored to different work environments.

This feature adjusts the level at which voices are received depending on the surrounding environment—such as the worksite, break room, office, or quiet meeting space—to ensure more accurate recognition of conversation content even in noisy conditions.

In addition, OTalk offers over 10 customized features, including one-on-one conversations, group chats, text translation, document translation, image translation, and real-time multilingual translation for training and lectures.

In particular, during safety training sessions for foreign workers, the instructor’s remarks are simultaneously translated into each worker’s native language. In group chats, content is translated according to each user’s language settings, facilitating work instructions and reports between on-site managers and foreign workers.

96% Adoption Rate Among Foreign Workers… Expansion to Partner Companies

On-site adoption is also rapidly increasing. Pilot program results showed that the O-Talk usage rate among foreign workers reached approximately 96%.

As requests from partner companies to use the solution continued to pour in, Hanwha Ocean and FLITTO Inc. expanded O-Talk’s coverage to include partner company workers after linking their employee ID numbers.

FLITTO Inc. plans to add support for languages such as Khmer and Sinhala in the future and further advance its text-to-speech technology. The company envisions expanding the solution’s scope beyond shipyards to high-risk industrial sites characterized by high noise levels and a large proportion of foreign workers.

Lee Jeong-soo, CEO of FLITTO Inc., stated, “This project, which overcame even dialects and slang, is an example of combining Hanwha Ocean’s accumulated experience in shipbuilding sites and language data with FLITTO Inc.’s AI translation and interpretation technology.” He added, “We will continue to expand the range of supported languages, such as Cambodian and Sri Lankan, and take the lead in breaking down language barriers at high-risk industrial sites by further advancing our text-to-speech technology.”