AI data and solutions company Flitto said on Tuesday it has built a customised AI interpretation and translation solution, OTalk, with Hanwha Ocean and Hanwha Systems for shipbuilding worksites.

OTalk is a solution planned by Hanwha Ocean's AI transformation (AX) division based on site requirements, and jointly built by Hanwha Systems and Flitto. It focuses on reducing communication problems caused by technical terms, slang and dialects at shipyards with many foreign workers.

Hanwha Ocean collected data on shipbuilding technical terms, slang and dialects used at worksites, and Flitto built a customised glossary and advanced its AI translation model based on it.

It trained voice data in seven major languages including Korean, English, Vietnamese and Thai, and improved recognition of site pronunciations mixed with Gyeongsang dialect accents, such as pronouncing "tank" as "ttaengk."

It applied a "five-level sound input adjustment" function that controls noise depending on the surrounding environment such as workshops, break rooms and offices. It also provides more than 10 functions including one-on-one conversations and group chats, text, document and image translation, and real-time multilingual translation for training and lectures.

In safety training for foreign workers, it enabled an instructor's speech to be delivered simultaneously in each worker's native language. In group chats, it translates work instructions from on-site managers and report content from workers according to each user's language settings.

During the pilot operation period, the usage rate among foreign workers was about 96 percent. It expanded the scope to cover workers at partner companies after linking employee numbers, in response to requests from partner companies to adopt it.

Flitto CEO Jeong-su Lee (이정수) said, "This project, which overcame even dialects and slang, is an example that combines Hanwha Ocean's accumulated experience and language data from shipbuilding worksites with Flitto's AI interpretation and translation technology." He added, "We will expand supported languages such as Cambodian and Sri Lankan and further advance voice synthesis technology."