SeaDoc: Visual Document Retrieval in Southeast Asian Languages

SeaDoc is a challenging visual document retrieval dataset for Southeast Asian languages, introduced in the paper 'Scaling Language-Centric Omnimodal Representation Learning'. It is designed to evaluate and enhance language-centric omnimodal embedding frameworks in low-resource settings. The dataset is hosted by LCO-Embedding and was last updated on 2026-02-06.

Use Cases

• Benchmarking visual document retrieval systems based on the described challenging task.
• Training language-centric omnimodal embedding models based on the described framework.
• Evaluating model performance on low-resource Southeast Asian languages based on the dataset's focus.

Strengths

• Designed for a specific, challenging task (visual document retrieval) as stated in the description.
• Focuses on low-resource Southeast Asian languages, addressing a noted research gap.
• Introduced in a peer-reviewed academic paper ('Scaling Language-Centric Omnimodal Representation Learning').

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Data may reflect geographic or linguistic bias inherent to its specific collection sources.

Provenance

Source

LCO-Embedding

Collection Method

Introduced in the academic paper 'Scaling Language-Centric Omnimodal Representation Learning'; specific collection method is not detailed in the provided input.

Freshness

Last updated 2026-02-06 10:51:16; freshness should be verified.

Geography

Southeast Asia (inferred from the focus on Southeast Asian languages)