A question-answering dataset about toponyms (place names) of Tatarstan, containing 38,696 QA pairs in Russian and Tatar languages. The dataset covers various aspects of geographical names including their type, coordinates, sources, etymology, administrative region, location, and physical characteristics.

Dataset Details

Dataset Description

This dataset provides extractive question-answering pairs about toponyms (place names) in Tatarstan, following the SQuAD format. It is designed to train and evaluate QA systems for geographical knowledge, specifically focused on Tatarstan's rich toponymic heritage. The dataset includes questions in both Russian and Tatar languages about various aspects of place names.

Curated by: TatarNLPWorld Community
Language(s): Russian (ru), Tatar (tt)
License: CC BY-SA 4.0
Format: SQuAD-style QA pairs
Domain: Toponymy, Geography, Linguistics

Dataset Sources

Repository: https://huggingface.co/datasets/TatarNLPWorld/tatarstan-toponyms-qa
Related Dataset: Tatarstan Toponyms

Uses

Direct Use

This dataset is intended for:

Training extractive question-answering systems
Evaluating QA models on geographical knowledge
Developing multilingual QA systems (Russian-Tatar)
Research on toponymy and geographical information retrieval
Educational applications for learning about Tatarstan geography

Out-of-Scope Use

This dataset should not be used for:

Open-domain QA without geographical focus
Tasks requiring real-time or dynamic geographical information
Generative QA without proper fact verification

Dataset Structure

Data Fields

Each record in the dataset follows the SQuAD format:

Field	Type	Description
`id`	`string`	Unique identifier for the QA pair
`title`	`string`	Title or name of the toponym
`context`	`string`	Text passage containing information about the toponym
`question`	`string`	Question about the toponym (in Russian or Tatar)
`answers`	`struct`	Object containing answer text and start position
`type`	`string`	Question type category
`subtype`	`string`	Question subtype (if applicable)

The answers struct contains:

text: The answer text extracted from the context
answer_start: Character position where the answer begins in the context

Data Statistics

General Statistics

Metric	Value
Total QA pairs	38,696
Question types	7
Languages	Russian, Tatar

Question Type Distribution

Type	Count	Percentage	Description
type	9,023	23.3%	Type/classification of the toponym
coordinates	9,023	23.3%	Geographical coordinates
sources	6,611	17.1%	Historical or documentary sources
etymology	5,707	14.7%	Origin and meaning of the name
region	4,289	11.1%	Administrative region or district
location	3,816	9.9%	Specific location description
phys	227	0.6%	Physical characteristics

Dataset Splits

Split	Size	Purpose
`train`	30,956	Training data for QA models
`validation`	3,870	Validation for hyperparameter tuning
`test`	3,870	Test set for final evaluation
`sample`	1,000	Small sample for quick experimentation

Example Entry

{
  "id": "toponym_1234_type",
  "title": "Казан",
  "context": "Казан (тат. Казан) — город в Российской Федерации, столица Республики Татарстан, крупный порт на левом берегу реки Волги при впадении в неё реки Казанки. Один из крупнейших экономических, научных, образовательных, религиозных, культурных и спортивных центров России. Казанский кремль является объектом Всемирного наследия ЮНЕСКО.",
  "question": "Какой тип объекта представляет Казань?",
  "answers": {
    "text": "город",
    "answer_start": 0
  },
  "type": "type",
  "subtype": ""
}

Tatar Language Example

{
  "id": "toponym_5678_etymology",
  "title": "Казан",
  "context": "Казан сүзенең килеп чыгышы турында берничә риваять бар. Бер риваять буенча, шәһәргә нигез салган хан алтын казанны Иделгә төшереп югалткан. Икенче версия буенча, 'казан' сүзе татар телендә 'каены' (каенлы урын) мәгънәсен аңлаткан 'казган' сүзеннән килеп чыккан.",
  "question": "Казан сүзенең этимологиясе нинди?",
  "answers": {
    "text": "берничә риваять бар",
    "answer_start": 24
  },
  "type": "etymology",
  "subtype": ""
}

Dataset Creation

Curation Rationale

The dataset was created to support question-answering research for low-resource languages and to preserve and make accessible knowledge about Tatarstan's geographical heritage. By providing QA pairs in both Russian and Tatar, we aim to:

Enable development of multilingual QA systems
Support educational applications for learning about Tatarstan
Preserve toponymic knowledge in digital form
Facilitate research in geographical information retrieval

Source Data

Data Collection and Processing

The dataset was constructed from structured toponymic data about Tatarstan:

Knowledge base: Created from the Tatarstan Toponyms dataset
Question generation: Template-based generation of questions for each toponym and attribute
Answer extraction: Automatic extraction of answer spans from context passages
Language expansion: Generation of questions in both Russian and Tatar
Quality validation: Manual review of sample QA pairs

Who are the source data producers?

The original toponymic data was compiled from various geographical, historical, and linguistic sources by the TatarNLPWorld community. The QA pairs were generated and validated by NLP researchers and native speakers of Russian and Tatar.

Annotations

Annotation process

The QA pairs were generated using:

Question templates: For each toponym attribute, multiple question templates were created
Automatic generation: Questions were generated for each toponym using the templates
Answer extraction: Answer spans were automatically identified in the context
Language adaptation: Questions were created in both Russian and Tatar
Quality control: Sample validation by native speakers

Who are the annotators?

The templates and validation were done by bilingual (Russian-Tatar) speakers from the TatarNLPWorld community.

Personal and Sensitive Information

The dataset contains only publicly available geographical information about places. No personal or sensitive information is included.

Bias, Risks, and Limitations

Known Limitations

Template bias: Questions are generated from templates, which may not cover all possible question formulations
Language coverage: Not all toponyms have Tatar language questions
Answer exhaustiveness: Answers are extracted from available context, which may not contain all possible information
Geographical coverage: Focuses on Tatarstan toponyms only
Temporal coverage: Information may not reflect the most current geographical data

Recommendations

Users should:

Consider the template-based nature when evaluating model generalization
Validate answers against authoritative sources for critical applications
Use the dataset for research and educational purposes
Supplement with additional data for production systems

Citation

BibTeX:

@dataset{tatarstan_toponyms_qa_2026,
    title = {Tatarstan Toponyms QA Dataset},
    author = {TatarNLPWorld Community},
    year = {2026},
    publisher = {Hugging Face},
    version = {1.0.0},
    url = {https://huggingface.co/datasets/TatarNLPWorld/tatarstan-toponyms-qa}
}

APA:

TatarNLPWorld Community. (2026). Tatarstan Toponyms QA Dataset [Data set]. Hugging Face. https://huggingface.co/datasets/TatarNLPWorld/tatarstan-toponyms-qa

Glossary

Toponym: A place name or geographical name
QA pair: Question-Answer pair
SQuAD format: Stanford Question Answering Dataset format with context, question, and answer spans
Etymology: The origin and historical development of a word/name

More Information

Version History

1.0.0 (2026-03-10): Initial release
- 38,696 QA pairs
- 7 question types
- Russian and Tatar languages

Dataset Card Authors

TatarNLPWorld Community

Dataset Card Contact

Hugging Face: https://huggingface.co/TatarNLPWorld

Part of the TatarNLPWorld resource ecosystem - Advancing Tatar Language Processing

Downloads last month: 13

Models trained or fine-tuned on TatarNLPWorld/tatarstan-toponyms-qa