Monolingual and Multilingual Misinformation Detection for Low-Resource Languages: A Comprehensive Survey

Most misinformation detection systems ignore the majority of world languages: The survey reveals that current approaches focus almost exclusively on high-resource languages (a handful of major world languages), leaving billions of speakers in low-resource language communities vulnerable to misinformation with minimal protective tools.

Data scarcity is a critical barrier to protecting non-English speakers: A fundamental challenge identified is the lack of labeled datasets for misinformation detection in low-resource languages, making it difficult to develop and test effective detection systems for these communities.

Cultural and linguistic context matters for accuracy: The survey emphasizes that misinformation detection cannot be one-size-fits-all; systems must account for cultural nuances and linguistic features specific to different languages and communities to be effective.

Emerging multilingual and multi-modal approaches show promise: The research identifies language-generalizable models and techniques that combine text with images or video as promising solutions that could extend misinformation detection capabilities across more languages simultaneously.

Responsible AI requires interdisciplinary collaboration and investment: The survey calls for stronger incentives for socially responsible AI research, improved data collection practices, and collaboration across disciplines to address the global misinformation challenge equitably across linguistic communities.