Kazakh News Corpus from 2020-2024 with 1.7 Million Tokens

1700 news publications in Kazakh, collected from major Kazakhstani news platforms like Tengri News and Egemen Kazakhstan between 2020 and 2024. The corpus contains 1,007,037 tokens, 107,501 types, and 109,395 lemmas, with a frequency list provided. It was compiled by Assel Ormanova and is available under a CC-BY-4.0 license.

Use Cases

• Train language models for Kazakh based on contemporary news text.
• Analyze linguistic patterns and vocabulary usage in modern Kazakh media.
• Conduct frequency analysis and build lexicons using the provided frequency list.
• Study news topics and discourse in Kazakhstan from 2020 to 2024.

Strengths

• Contains 1,007,037 tokens from a defined 5-year period (2020-2024).
• Sourced from 12+ named, major Kazakhstani news platforms, suggesting diverse journalistic content.
• Includes a frequency list and lemma counts (109,395 lemmas) for linguistic analysis.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• The 4.0 MB file size indicates a relatively small corpus for modern NLP training.
• Data may reflect the editorial bias inherent to the selected news sources.

Provenance

Source

Compilation from Kazakhstani news platforms (e.g., Tengri News, Egemen Kazakhstan).

Collection Method

Downloaded from news platforms and processed with #LancsBox software.

Time Range

2020-2024

Freshness

Last updated 2026-05-28 11:57:08; freshness should be verified.

Geography

Kazakhstan