Unga: UN General Assembly Debate Speeches for Thematic Classification

A 2026 paper by Mumtaz et al. introduces this dataset for evaluating the temporal robustness of large language models. The dataset likely contains transcripts from United Nations General Assembly debates, annotated for thematic classification. It was uploaded to Hugging Face by the author SLPG on June 10, 2026.

Use Cases

• Benchmarking LLM robustness over time based on UN debate transcripts mentioned in the description
• Training models for thematic classification of political speech based on the described application
• Studying temporal drift in language and topics within international diplomatic discourse
• Evaluating model generalization on historical versus contemporary text data

Strengths

• Associated with a peer-reviewed 2026 publication in the journal 'Information'.
• Specifically designed for evaluating temporal robustness, a key challenge in NLP.

Limitations

• Description metadata is limited; actual data quality requires manual inspection after download.
• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.

Provenance

Source

United Nations General Assembly debates, as inferred from the associated paper's title.

Collection Method

Likely collected and annotated from official UN transcripts for research purposes.

Freshness

Last updated 2026-06-10 12:06:37; freshness should be verified.

Geography

Global, as inferred from the UN General Assembly source.