Page 442 - AI for Good Innovate for Impact

AI for Good Innovate for Impact

• The use case addresses a critical need for innovation in the construction sector. According
to the Productivity Statistics Portal of the Korea Productivity Center (KPC) [2], South Korea’s
per capita labor productivity index declined in 2022... While productivity in most industries
has improved, the construction sector has continued to fall behind, underscoring the
need for technological interventions like AI-driven translation solutions. In this context,
the development of an AI-powered interpretation app trained in construction-specific
terminology represents a groundbreaking innovation in the sector.

By reducing language barriers, improving safety, and enhancing workflow clarity, this AI-
powered app directly contributes to productivity gains and safer working environments—
making it a valuable innovation for the construction and shipbuilding sectors.

2�3 Future Work

We will continuously collect newly emerging technical terms and on-site slang frequently used
in construction sites to improve translation accuracy. In addition, we aim to further enhance the
AI model to enable more accurate speech recognition even in noisy environments.

Furthermore, we plan to develop additional features that provide safety training and manuals
for foreign workers, thereby contributing to the prevention of serious industrial accidents and
promoting a culture of safety across the construction industry.

3 Use Case Requirements
• REQ-01: It is critical that effective communication for safety is ensured at construction
sites with a high proportion of foreign workers. Language barriers in such environments
can significantly impede communication and increase the likelihood of serious industrial
accidents.
• REQ-02: It is critical to build and train AI models using domain-specific datasets that
include construction-specific terminology and slang. Techniques such as fine-tuning and
prompt engineering must be applied to improve translation accuracy.
• REQ-03: It is critical to validate the model’s real-world applicability. Field testing was
conducted in more than 20 languages, including English, Chinese, Hindi, Thai, and
Vietnamese, achieving over 90% user satisfaction.
• REQ-04: It is critical that the app performs reliably in noisy construction environments. To
this end, STT (Speech-to-Text) and TTS (Text-to-Speech) capabilities were incorporated
to enhance voice recognition accuracy and improve user experience.

406