sabin1234/SAC_Nepal_FAQ_Nepali_Health_Fitness_Dataset

SAC Nepal FAQ — Nepali Health & Fitness Dataset

Overview

This dataset (sac_nepal_faq_nepali.jsonl) is a collection of 100 instruction-following conversation pairs in Nepali, covering frequently asked questions about health, fitness, and nutrition. Each record is a single-turn human↔gpt exchange: a Nepali-language question followed by an informative Nepali-language answer.

The data appears to be a localized/translated set — the behavior_definition field for every record states that answers are produced by "converting the original question's meaning and the original answer's facts into pure Nepali" (i.e., the underlying Q&A pairs were adapted into Nepali while preserving the source facts).

• —File format: JSON Lines (.jsonl), one JSON object per line
• —Total records: 100
• —Source: SAC Nepal (source_name: sac_nepal_faq)
• —License: Apache-2.0 (permissive tier)
• —Language: Nepali (ne / ISO 639-3 npi), Devanagari script (Deva)
• —Task type: instruction-following
• —Generation type / condition: real (i.e., real rather than synthetic data)

File / Record Structure

Every line is a JSON object with the following top-level fields:

metadata_json sub-fields (per record)

Distribution Analysis

Behavior Distribution

There is only one behavior type across the entire dataset:

All 100 records share the exact same behavior_definition string, meaning the dataset was generated under a single, uniform instruction: translate/adapt an existing Q&A pair into faithful, fact-preserving Nepali.

Question Type Distribution

Every question in the dataset is fact-based (as opposed to opinion-based, hypothetical, or open-ended creative prompts).

Domain / Category / Sub-domain Distribution

The dataset is single-domain (Health) but internally spans several health sub-topics, roughly clustered in this order as the records progress from 001 to 100:

1. 1.General fitness fundamentals (~#1–10): importance of health/fitness, how to start a routine, exercise frequency, benefits of physical activity, risks of inactivity
2. 2.Exercise types & routines (~#11–30): warm-ups, best workout times, types of exercise, building a balanced/home workout plan, strength training, fat-burning
3. 3.Diet & nutrition (~#21–40): balanced diet, macronutrients, healthy eating habits
4. 4.Weight management (~#31–40): sustainable weight loss, healthy weight gain
5. 5.Heart health (~#41–50): cardiovascular exercise, improving heart health
6. 6.Mental health (~#51–60): exercise's effect on depression/anxiety, stress management
7. 7.Injury prevention & equipment (~#61–80): avoiding common workout injuries, home workout equipment needs
8. 8.Special populations (~#81–90): fitness for senior citizens, and likely other groups
9. 9.Supplements & performance (~#91–100): pre-workout supplements, creatine, testosterone, protein, caffeine, hydration/electrolytes, performance-enhancing drug risks, weight-loss supplements

(Note: the structured `generation_sub_domain` metadata field is identical across all 100 rows — the topic progression above is inferred from reading the actual question text, since the topical variety lives in the content itself rather than in separate metadata tags per sub-topic.)

Content Language / Script Distribution

No code-switching or English content is permitted anywhere in this dataset — it is intended to be purely monolingual Nepali in Devanagari script.

Other Constant Fields

The following fields are identical across all 100 records (i.e., no variation to report): source, source_name, source_repo, source_config, source_split, source_revision, language, language_code, script, license, license_tier, task_type, generation_type, condition, and url (always empty).

Conversation Structure

This is a single-turn instruction/response dataset — there is no multi-turn dialogue or follow-up questions.

Text Length Statistics

Questions are short, single-sentence prompts; answers are concise (roughly 1–3 sentences), informative responses — consistent with an FAQ-style format rather than long-form explanation.

Summary

This dataset is a narrow, single-purpose FAQ corpus: 100 short Nepali-language question/answer pairs, all in the health & fitness domain, all following one uniform "informative, fact-preserving Nepali translation" behavior, and all classified as fact-based questions. There is no variation across behavior type, question type, domain, license, or task type — the only meaningful variety in the dataset is in the actual subject matter of the questions themselves, which progress through general fitness, exercise types, diet, weight management, heart health, mental health, injury prevention, special populations, and dietary supplements.

This kind of dataset is typically useful for fine-tuning or evaluating a language model's ability to answer health/fitness FAQs fluently and accurately in Nepali, or as seed data for building a larger Nepali-language instruction-tuning corpus.