This dataset contains posts and comments from the VK community "Аллахтан" (Allahtan), a Tatar-language community focused on Islamic spirituality, religious education, moral discussions, and cultural content related to Islam in the Tatar tradition. The data was collected to support NLP research and development for the Tatar language in the religious and spiritual domain, as well as for tasks such as text classification, sentiment analysis, religious discourse analysis, and social media analysis.

Curated by: TatarNLPWorld
Language(s): Tatar (mainly), Russian (some code-switching, especially in religious terminology)
License: Apache 2.0

Dataset Sources

Repository: TatarNLPWorld/allahtan-vk

Uses

Direct Use

This dataset can be used for:

Religious text classification
Sentiment analysis in spiritual contexts
Discourse analysis of religious discussions
Language modeling for Tatar (religious domain)
Code-switching detection in religious contexts
Social media analysis of faith-based communities
Islamic terminology extraction
Moral and ethical discussion analysis
Tatar Islamic literature studies

Out-of-Scope Use

This dataset should not be used for:

Identifying individual users
Religious profiling or discrimination
Surveillance or monitoring of religious groups
Any purpose that violates VK's terms of service
Misrepresentation of Islamic teachings

How to Use

from datasets import load_dataset

# Load posts
posts = load_dataset("TatarNLPWorld/allahtan-vk", "posts", split="full")
posts_sample = load_dataset("TatarNLPWorld/allahtan-vk", "posts", split="sample")

# Load comments
comments = load_dataset("TatarNLPWorld/allahtan-vk", "comments", split="full")
comments_sample = load_dataset("TatarNLPWorld/allahtan-vk", "comments", split="sample")

# Convert to pandas for analysis
posts_df = posts.to_pandas()
comments_df = comments.to_pandas()

# Merge posts and comments
merged = posts_df.merge(comments_df, on='post_id', how='left')

# Example: Show first few rows
print(posts_df.head())
print(comments_df.head())

# Example: Get basic statistics
print(f"Total posts: {len(posts_df)}")
print(f"Total comments: {len(comments_df)}")
print(f"Average post length: {posts_df['content'].str.len().mean():.1f} characters")
print(f"Average comment length: {comments_df['content'].str.len().mean():.1f} characters")

# Example: Find most engaging religious discussions
top_discussions = posts_df.nlargest(10, 'comments_count')[['content', 'comments_count', 'likes_count']]
print(top_discussions)

Dataset Structure

Data Instances

Posts:

{
  "post_id": 1090676,
  "owner_id": -92329133,
  "url": "https://vk.com/wall-92329133_1090676",
  "date": "2026-03-10 11:34:10",
  "content": "Аллаһы Тәгаләнең рәхмәте... (The mercy of Allah Almighty...)",
  "author_id": -92329133,
  "likes_count": 3584,
  "comments_count": 694,
  "reposts_count": 3292,
  "views_count": 195117
}

Comments:

{
  "comment_id": 1090697,
  "post_id": 1090658,
  "owner_id": -92329133,
  "parent_comment_id": 0,
  "from_id": 1078645425,
  "date": "2026-03-10 11:34:10",
  "content": "Амин! Рәхмәт сезгә... (Amin! Thank you...)",
  "likes_count": 66
}

Data Fields

Posts:

post_id: Unique identifier for the post
owner_id: ID of the community/page that owns the post (-92329133 for this community)
url: Direct URL to the post
date: Post creation timestamp (UTC)
content: Text content of the post (religious teachings, Quranic verses, hadith, moral discussions)
author_id: ID of the post author
likes_count: Number of likes on the post
comments_count: Number of comments on the post
reposts_count: Number of reposts
views_count: Number of views

Comments:

comment_id: Unique identifier for the comment
post_id: ID of the post this comment belongs to
owner_id: ID of the community/page
parent_comment_id: ID of parent comment (always 0 in this dataset, no nested replies)
from_id: ID of the comment author
date: Comment creation timestamp (UTC)
content: Text content of the comment
likes_count: Number of likes on the comment

Data Splits

The dataset is split into two configurations:

posts: Contains only posts
comments: Contains only comments

Each configuration has two splits:

full: Complete dataset
sample: 1000 random examples for quick experimentation

Dataset Creation

Curation Rationale

The dataset was created to address the lack of publicly available Tatar language data in the religious domain. The Allahtan community was chosen because:

It's one of the largest Tatar-language religious communities
Long historical span (2015-2026) showing evolution of online religious discourse
High engagement from the Tatar-speaking Muslim community
Rich vocabulary of Islamic terminology in Tatar
Mix of educational content and community discussions

Source Data

Data Collection and Processing

Data was collected from the VK community "Аллахтан" (https://vk.com/allahtan) using the official VK API. The collection process:

All posts from the community were collected (18,844 posts)
All comments on those posts were collected (85,761 comments)
Data was cleaned and formatted into CSV format
Personal identifiers were preserved for research purposes but should be handled responsibly

Who are the source data producers?

The data was produced by members of the Allahtan VK community, including:

Religious educators and imams
Practicing Muslims sharing knowledge
Community members seeking spiritual guidance
People discussing moral and ethical questions
Those interested in Tatar Islamic traditions

Annotations

The dataset does not contain manual annotations. All data is raw text from VK with associated metadata. However, the religious nature of the content provides natural categories that could be used for:

Topic classification (Quranic verses, hadith, moral advice, prayers)
Sentiment analysis in religious context
Question-answer pairs (religious consultations)
Religious terminology extraction

Personal and Sensitive Information

The dataset contains:

User IDs (VK internal identifiers)
Public comments and posts about religion
Public interaction counts (likes, reposts, views)

⚠️ Sensitive Content Notice: This dataset contains religious discussions that some users may consider sensitive. It includes:

Discussions of Islamic beliefs and practices
Moral and ethical guidance
Religious interpretations
Personal spiritual experiences

No private messages, email addresses, phone numbers, or other personally identifiable information is included. However, user IDs could potentially be used to cross-reference with other public data. Researchers should handle this data responsibly, with respect for religious sensitivity, and not attempt to de-anonymize users.

Bias, Risks, and Limitations

Known Biases

Religious bias: Data represents Islamic discourse from a specific Tatar community, not all religious perspectives
Community bias: May reflect particular interpretations of Islam
Platform bias: Social media language differs from formal religious texts
Temporal bias: Posts span from 2015 to 2026 (3,917 days), showing evolution in online religious discourse
Engagement bias: More active users may be overrepresented
Theological bias: May not represent all schools of Islamic thought

Content Limitations

Contains religious terminology requiring domain knowledge
Code-switching between Tatar and Russian, especially in religious concepts
Includes prayers, supplications, and religious poetry
May contain discussions of sensitive religious topics
Comments are all top-level (no nested replies)
Some content may be considered sacred by believers

Recommendations

Users should:

Approach religious content with respect and cultural sensitivity
Be aware of potential theological biases
Consult with domain experts when interpreting religious content
Not use for religious profiling or discrimination
Handle user IDs with care and respect user privacy
Consider the context of Islamic discourse in Tatar culture
Be mindful that translations of religious terms may be sensitive

Dataset Statistics

Posts Statistics

Total posts: 18,844
Date range: 2015-06-19 to 2026-03-10 (3,917 days / ~10.7 years)
Total characters: 5,856,933
Average post length: 311 characters
Total likes: 553,851 (avg 29.4 per post)
Total comments: 234,556 (avg 12.5 per post)
Total reposts: 232,616 (avg 12.3 per post)
Total views: 68,213,389 (avg 3,620 per post)
Posts with comments: 15,908 (84.4%)
Posts without comments: 2,936 (15.6%)
Posts with likes: 18,414 (97.7%)
Posts without likes: 430 (2.3%)
Posts with reposts: 11,192 (59.4%)
Posts without reposts: 7,652 (40.6%)
Most liked post: 3,584 likes
Most commented post: 694 comments
Most viewed post: 195,117 views
Most reposted post: 3,292 reposts

Comments Statistics

Total comments: 85,761
Total characters: 5,171,134
Average comment length: 60 characters
Total likes on comments: 140,222 (avg 1.6 per comment)
Unique comment authors: 83,146 (96.95% of comments are from unique authors)
Top-level comments: 85,761 (100%, no nested replies)
Comments with likes: 54,765 (63.9%)
Comments without likes: 30,996 (36.1%)
Most liked comment: 66 likes

Overall Dataset

Total records: 104,605
Total characters: 11,028,067
Total size: ~26 MB (raw CSV)
Total interactions (likes + comments + reposts): 1,161,245

Temporal Analysis

Longest running community: 10.7 years of activity
Posts per year (average): ~1,760 posts/year
Comments per year (average): ~8,010 comments/year
Growth trend: Steady engagement from 2015 to 2026

Engagement Analysis

Comment-to-post ratio: 12.5 comments per post (high engagement)
Like-to-post ratio: 29.4 likes per post
Repost-to-post ratio: 12.3 reposts per post
Comment like rate: 1.6 likes per comment (moderate engagement)
Viral posts: Multiple posts with >3,000 reposts and >100,000 views

Citation

If you use this dataset in your research, please cite:

@dataset{allahtan-vk-2026,
  title = {Allahtan VK Dataset: Tatar Language Religious and Spiritual Posts and Comments},
  author = {Arabov Mullosharaf Kurbonovich},
  year = {2026},
  publisher = {Hugging Face},
  url = {https://huggingface.co/datasets/TatarNLPWorld/allahtan-vk}
}

Dataset Card Authors

TatarNLPWorld Team

Dataset Card Contact

For questions or feedback, please contact through the Hugging Face repository issues page.

Downloads last month: 1

Total file size:

15.3 MB