salyamq/kbtu-qa

KBTU Student FAQ Dataset

Dataset Description

This dataset contains 190 frequently asked questions and answers collected from KBTU (Kazakh-British Technical University) students, covering everyday academic and administrative topics: exams, retakes, scholarships, syllabi, certificates, clubs, lost & found, and other common student concerns.

Each entry is provided in both Russian (original) and English (translated), along with the original unedited answer for transparency.

• Language(s): Russian (ru), English (en)
• License: MIT

Dataset Structure

Data Fields

Field	Type	Description
id	int	Unique identifier (1–190)
question	string	Original question, in Russian, as written by students
answer	string	Grammar-corrected answer in Russian (meaning preserved, only language cleaned up)
original_answer	string	Raw, unedited answer in Russian, exactly as originally written
question_eng	string	English translation of the question
answer_eng	string	English translation of the corrected answer

Data Instance Example

{
  "id": 14,
  "question": "Когда выплачивают стипендию и в каких числах месяца она приходит?",
  "answer": "Никто не знает, когда именно придёт стипендия, но обычно после 20-го числа.",
  "original_answer": "Никто не знает когда именно придет стипендия, но обычно после 20-го числа",
  "question_eng": "When is the scholarship paid, and on what dates does it usually arrive?",
  "answer_eng": "Nobody knows exactly when the scholarship will arrive, but it's usually after the 20th."
}

Data Splits

The dataset consists of a single split with 190 examples.

Dataset Creation

Source Data

Questions and answers were originally collected and written manually by students/staff based on real, recurring FAQ topics at KBTU.

Processing

1. Collection: Raw Q&A pairs compiled in Markdown format from real student inquiries.
2. Structuring: Converted to structured JSON (id, question, answer) with full content-integrity validation (character-level comparison against the source to ensure nothing was lost or altered).
3. Grammar correction: The answer field was passed through an LLM (DeepSeek V4) with a strict system prompt instructing it to fix grammar, spelling, and punctuation only — no rephrasing, no added or removed information. The original text is preserved separately in original_answer for full traceability.
4. Translation: The (corrected) Russian question/answer pairs were translated into English (question_eng, answer_eng) using DeepSeek V4, with a glossary of KBTU/university-specific terms enforced for consistency (e.g. "ретейк" → "retake", "WSP" left untranslated).

Annotations

No manual annotation beyond grammar correction and translation (both LLM-assisted, described above). No human relabeling of content or intent.

Considerations for Using the Data

• Domain-specific: Answers reference KBTU-specific systems, offices, and terminology (e.g. WSP, UniX, deanship offices) and may not generalize to other universities without adaptation.
• LLM-assisted correction & translation: While the correction step was constrained to preserve meaning, and translations followed a fixed terminology glossary, this data has not been manually reviewed line-by-line post-processing. Users requiring perfect fidelity should spot-check against original_answer.
• Informal register: Some answers include informal, colloquial phrasing (e.g. "nobody really knows exactly when..."), reflecting real student-authored FAQ content rather than official university documentation.

Citation

If you use this dataset, please cite it as:

@misc{salim_tokhtobayev_2026,
    author       = { Salim Tokhtobayev },
    title        = { kbtu-qa (Revision 0fe4ac0) },
    year         = 2026,
    url          = { https://huggingface.co/datasets/salyamq/kbtu-qa },
    doi          = { 10.57967/hf/9726 },
    publisher    = { Hugging Face }
}