🇸🇬Singapore·AI News·16 Nov 2026·via E27

    The Philippines can’t win the AI arms race, so it is betting on Taglish

    AI models trained on Western internet data struggle to understand code-switched languages like Taglish, a blend of Tagalog and English prevalent in the Philippines. This linguistic gap means that while a human Manila contact-centre agent can instantly interpret a phrase like 'Pakicheck naman po ang bill ko, may extra charge yata,' a standard AI model may misinterpret the query.

    The challenge for artificial intelligence in the Philippines is its ability to process Taglish, a common linguistic practice where Tagalog and English are intermixed within a single conversation or sentence. A Manila contact-centre agent can immediately grasp the meaning of a Taglish phrase from a caller and respond before the sentence concludes.

    However, an AI model trained predominantly on data from Silicon Valley's internet, which largely consists of English, may struggle with this specific type of code-switching. Such models are prone to misinterpreting queries that blend Tagalog and English, potentially leading to inaccurate or irrelevant responses.

    This discrepancy highlights a practical limitation for AI products in linguistically diverse markets. While mainstream AI tools can process standard English, their effectiveness diminishes when precise understanding of localized, mixed languages like Taglish is required. For businesses relying on AI in customer service or other communication-heavy roles in the Philippines, this gap necessitates specialized development or training to achieve human-level comprehension.

    Share this article

    Go deeper
    Original reporting by E27We don't republish, read the full story →

    Related reading

    6 stories