Setswana Grammar Checker Corpus: 1,700 Annotated Sentences for NLP

Irina Zlotnikova developed a manually curated and annotated corpus of 1,700 Setswana sentences for training a grammar checking system. The dataset supports a Long Short-Term Memory (LSTM) model that achieved 96% classification accuracy for grammatical correctness. The work was published on figshare in April 2026.

Use Cases

• Training grammar checking models for Setswana based on the annotated sentence corpus.
• Evaluating neural network performance for applied linguistics tasks in low-resource settings.
• Supporting grammar teaching and educational technology development for Setswana speakers.
• Providing a foundational resource for morphosyntactic pattern analysis in Southern Bantu languages.

Strengths

• Manually curated and annotated corpus of 1,700 Setswana sentences.
• Model trained on the data achieved a classification accuracy of 96%.
• Dataset is openly licensed under CC-BY-4.0.

Limitations

• Row count is unknown, which may limit suitability assessment.
• Column-level documentation is absent; field semantics must be inferred after download.
• The dataset is very small at 18.6 KB, indicating a limited scope.

Provenance

Source

Irina Zlotnikova via figshare.

Collection Method

Manually curated and annotated corpus.

Freshness

Last updated 2026-04-10 11:38:43; freshness should be verified.

Geography

Setswana language, spoken in Southern Africa (e.g., Botswana, South Africa).