Singapore launches Singlish-speaking AI to assist seniors, block scam calls

Developed by the A*Star Institute, Meralion was first made available as a free download in December 2024. The latest version can process at least eight regional languages, including English, Mandarin, Tamil, Malay, Thai, and Singlish, the city-state’s distinctive language, as reported by The Straits Times. Developers said support for Chinese dialects will be added in future updates.

It is part of a SGD70 million (over US$54 million) initiative funded by the National Research Foundation and the Infocomm Media Development Authority to develop large language models tailored for Southeast Asia.

Speaking at the Asia Tech x Singapore conference on May 28, Digital Development and Information Minister Josephine Teo highlighted Meralion’s ability to interpret non-verbal cues such as volume, tone, and emotion. "I’ve been told (the program) can also handle non-verbal cues such as the speaker’s volume, emotion and tone," she said.

Singapore's iconic Merlion. Photo from Pexels

Meralion aims to address the limitations of widely used AI models, which are often trained primarily on Western data and struggle with regional dialects, linguistic patterns, and communication styles. Trained on Singapore’s national speech corpus, it is designed to recognize multilingual sentences, reflecting how people in the region naturally communicate.

"It’s very unusual for us to complete a whole sentence using just one language," Teo said, noting that Southeast Asia is home to more than 1,200 languages and dialects. She added that Meralion could potentially serve over 450 million people in the region who use these languages.

Axiom IT Solutions, which is working with Meralion to develop new applications, is currently in talks with a social service agency to deploy the technology in eldercare. The AI model can autonomously call seniors, remind them to take medication, and assess their emotional well-being by analyzing speech patterns for signs of distress. After each interaction, the system generates a summary report to help social workers track individual needs.

It is also being tested to screen scam calls, identifying itself to suspicious callers and asking their purpose before deciding whether to allow or block the call. It is also capable of filtering out automated bot calls, commonly used in mass-targeting scams.

Lam Pang Ngean, business development director at Axiom, said telcos may consider using the system to screen suspicious calls before they reach users who opt in, or as a downloadable app for personal call filtering.