AI Support for Indonesian vs. Malaysian Languages

Uncover the differences in AI support for Indonesian and Malaysian languages, focusing on market size, training data, and technological priorities.

Artificial Intelligence (AI) is rapidly advancing globally, connecting nations and cultures like never before. Throughout Southeast Asia, AI technology has become a significant force for digital transformation—especially in language processing and understanding. Two closely related languages—Indonesian (Bahasa Indonesia) and Malaysian (Bahasa Malaysia)—are increasingly benefiting from AI advancements. But does AI favor one language more than the other?

Understanding the Malaysian and Indonesian Language Landscape

Both Bahasa Malaysia and Bahasa Indonesia share common roots, with mutual intelligibility between speakers of both languages. However, when it comes to artificial intelligence systems, there's a notable disparity in support levels that affects millions of speakers across Southeast Asia. In this blog post, we will attempt to closely analyze whether Indonesian language presently holds more AI support than Malaysian, and identify key factors driving this situation.

Historical Context and Linguistic Relationship

Bahasa Malaysia and Bahasa Indonesia originate from Malay but evolved separately due to colonial influences and national standardization efforts. According to linguistic studies quoted in Wikipedia, they share approximately 85% lexical similarity, making them mutually comprehensible despite their differences (Asmah, 2019).

Despite this close relationship, AI systems often treat these languages differently, with implications for speakers across both countries.

Factors Driving Greater AI Support for Indonesian

Population and Market Size Disparities

The most significant factor explaining the support gap is simple demographics:

• Indonesia: 281.2 million people (World Bank, 2025)
• Malaysia: 35.1 million people (World Bank, 2025)

This population differential creates a substantially larger market for language technologies, influencing development priorities. According to research by Statista's Digital Market Outlook, Indonesia represents a digital market approximately four times larger than Malaysia's.

Digital Content Availability and Training Data

AI language models rely heavily on training data, and the disparity in available digital content is substantial:

• Indonesian content represents approximately 0.4% of Common Crawl data
• Malaysian language content constitutes just 0.09% of the same dataset

This means Indonesian has over four times the representation in this crucial training resource (Joshi et al., 2022).

Wikipedia statistics further illustrate this gap:

• Indonesian Wikipedia: ~590,000 articles
• Malaysian Wikipedia: ~380,000 articles

Commercial Prioritization by Tech Companies

Major AI providers typically implement support for languages with larger speaker populations first:

• Google Cloud Translation API implemented Indonesian support in 2016, with Malaysian added later in 2019
• Microsoft Azure Cognitive Services followed a similar pattern
• OpenAI's models demonstrate stronger performance on Indonesian tasks compared to Malaysian ones

Measuring AI Support: A Comparative Analysis

Performance in Major Language Models

Recent benchmarks reveal consistent performance disparities:

Accuracy measured on composite tasks including classification, summarization, and question answering (FLORES-200 Benchmark, 2023)

Translation Quality Assessment

The FLORES-200 benchmark for machine translation quality shows that English-to-Indonesian translation consistently outperforms English-to-Malaysian translation:

• Google Translate achieves a BLEU score of 42.6 for Indonesian versus 38.7 for Malaysian
• DeepL reports similar disparities with 41.3 versus 36.8 respectively

These differences directly impact user experience for speakers of both languages.

The Current State of Malaysian Language in AI

Despite receiving less attention than Indonesian, Malaysian language support is improving:

• BERT4Malay: A Malaysian language BERT model released in 2021 achieved significant improvements for downstream NLP tasks
• MalayaKT: An open-source toolkit specifically for Malaysian language processing
• BM-BERT: A specialized model outperforming multilingual BERT on Malaysian-specific tasks

Bridging the Gap: Future Outlook

Recent Initiatives

Several initiatives are actively working to improve Malaysian language support:

• The Malaysian government's National Language Technology Initiative, launched in 2022
• MIMOS (Malaysia's national R&D center) collaborations with universities
• The Malaysian Natural Language Processing Association formed in 2021

Cross-Lingual Approaches

Promising developments that may narrow the gap include:

• Cross-lingual transfer learning techniques leveraging Indonesian resources to improve Malaysian language processing
• Joint Indonesian-Malaysian models showing better performance than language-specific ones in some applications
• Parameter-efficient fine-tuning methods making it more economical to adapt existing models to Malaysian

Practical Implications and Recommendations

For Malaysian Language Users

Malaysian language users can improve their AI experiences by:

• Using Indonesian as a proxy language for certain applications while being mindful of vocabulary differences
• Contributing to open-source Malaysian language datasets
• Providing feedback to technology companies about Malaysian language needs

For Developers

For those developing applications for Malaysian-speaking users:

• Consider hybrid models leveraging both Indonesian and Malaysian training data
• Implement dialect detection to dynamically adjust processing
• Test applications specifically with Malaysian users rather than assuming Indonesian performance will transfer

Conclusion

The disparity in AI support between Indonesian and Malaysian languages stems from market size, digital content availability, and commercial prioritization. While Indonesian currently enjoys broader implementation, Malaysian language support is improving through dedicated research and growing awareness of linguistic diversity.

As AI development continues to emphasize inclusivity and techniques for efficient adaptation to lower-resource languages mature, we can expect the gap between these closely related languages to narrow. Both users and developers can take specific actions to improve Malaysian language experiences while contributing to the broader ecosystem of language technology in Southeast Asia.