AI Tech / news
AI Developers Take On Cantonese and Hokkien Language Gaps
Chinese dialects such as Cantonese and Hokkien are less widely spoken than in the past, and mainstream AI tools often struggle with their local vocabulary and colloquial grammar. Hong Kong-based developers including Votee AI are building models for Cantonese and other underserved languages.
Families across Southeast Asia can include grandparents fluent in Hokkien and Cantonese but only conversational in Mandarin, leaving many home languages largely absent from digital tools. That mismatch matters as AI becomes more embedded in daily life, because the technology’s reach may shape which languages are passed down to future generations.
Mainstream AI systems are not incapable of handling Chinese dialects. Their performance varies, and some can manage everyday conversations reasonably well. But familiarity with a language is not the same as command of local vocabulary, cultural references or colloquial grammar.
As a result, a model that works well in Mandarin can still misunderstand a Cantonese expression or struggle with a Hokkien conversation. The shortfall is less about basic comprehension than about the specific textures of speech that families actually use.
Developers are beginning to address the gap. Hong Kong startups are training models specifically for Cantonese, while researchers are building speech technologies for Hokkien.
Votee AI, a Hong Kong-based company, is developing language models focused on Cantonese and other languages underserved by mainstream AI. Its work is part of a broader effort to make AI more useful for dialects that remain central to family life but are less visible online.