Kazakh Human Curated Instruction Dataset
Collection
The dataset is human curated instruction dataset for fine-tuning LLM. • 29 items • Updated
This dataset contains 1,504 samples focusing on common orthographic and grammatical nuances in the Kazakh language.
| Metric | Count |
|---|---|
| Total Samples | 1,504 |
| Total Words (approx.) | 68,469 |
| Avg. Words per Sample | 45 |
The following table details the distribution of word counts across different fields:
| Field | Mean | Median | Min | Max | Total Words |
|---|---|---|---|---|---|
| correction | 22.8 | 22.0 | 8 | 89 | 34,308 |
| domain | 1.2 | 1.0 | 1 | 4 | 1,764 |
| prompt | 8.6 | 8.0 | 3 | 18 | 12,912 |
| response | 6.0 | 6.0 | 2 | 17 | 8,982 |
| text | 6.0 | 6.0 | 2 | 17 | 8,999 |
sample_id (int64): Unique numerical identifier.prompt (string): The user request (e.g., "Check for orthographic and grammatical errors").text (string): The input sentence containing a subtle error.response (string): The linguistically accurate version of the sentence.correction (string): A rule-based explanation (e.g., explaining why plural suffixes are omitted after quantifiers like "көп").domain (string): The contextual domain, including Business and General communication. {
"sample_id": 1,
"prompt": "Мәтіндегі орфографиялық және грамматикалық қателерді тексеріп беріңіз.",
"text": "Бүгін дүкеннен көп алмалар сатып алдым.",
"response": "Бүгін дүкеннен көп алма сатып алдым.",
"correction": "Қазақ тілінде \"көп\", \"бірнеше\" сияқты мөлшерлік сөздер зат есімнің алдына келгенде, зат есімге көптік жалғауы жалғанбайды. Себебі мөлшерлік сөздің өзі көптікті білдіреді.",
"domain": "Business"
}