OiQ/fikran-dataset

Fikran Conversations Dataset

Complete dataset of conversations from Fikran, an Arabic social platform focused on AI and intellectual discussions.

Dataset Overview

• Platform: Fikran (fikran.com)
• Content Type: Blog posts with full metadata
• Language: Arabic (primary), English (secondary)
• Total Samples: 373 conversations
• Data Format: JSONL (JSON Lines)
• License: CC BY-NC-SA 4.0

Dataset Contents

• Blog Posts: 373 posts with complete content
• User Information: Author profiles with metadata
• Categories: Blog categories
• Tags: Hashtags
• Reactions: 6 reaction types (like, love, haha, wow, sad, angry)
• Comments: Comment threads structure (nested replies)
• Timestamps: Publication dates in Arabic format

Dataset Structure

The dataset is provided in JSONL format, with one JSON object per line.

File Organization

fikran-dataset/
├── README.md                    # This file
├── data/
│   ├── train.jsonl            # Training data (300 samples)
│   ├── test.jsonl             # Test data (73 samples)
│   └── blogs.jsonl            # All blog posts (373 samples)
└── dataset_infos.json          # Dataset metadata

Data Fields

Each conversation/post contains:

Field	Type	Description
id	string	Unique conversation identifier
platform	string	Platform name ("fikran")
type	string	Content type ("blog")
title	string	Post title
content	string	Full post content
url	string	Original post URL
language	string	Content language ("ar", "en")
author	object	Author information
author.username	string	Author username
author.display_name	string	Author display name
author.profile_url	string	Author profile URL
timestamp	string	Publication timestamp
category	object	Category information
tags	list	Hashtags
stats	object	Engagement statistics
stats.views	integer	View count
stats.likes	integer	Like count
stats.reactions	object	Reaction counts by type
comments	list	Comment threads

Dataset Splits

The dataset includes the following splits:

Split	Samples	Description
train	300	Training data (first 300 posts)
test	73	Test data (last 73 posts)
blogs	373	All blog posts (no split)

Usage

Loading with Datasets Library

from datasets import load_dataset

# Load all data
dataset = load_dataset("OiQ/fikran-dataset")

# Load specific split
train_data = load_dataset("OiQ/fikran-dataset", split="train")
test_data = load_dataset("OiQ/fikran-dataset", split="test")

# Access data
for example in dataset["train"]:
    print(f"Title: {example['title']}")
    print(f"Content: {example['content']}")
    print(f"Author: {example['author']['username']}")

Loading with Pandas

import pandas as pd

# Read JSONL file
df = pd.read_json('data/blogs.jsonl', lines=True)

# Basic statistics
print(f"Total posts: {len(df)}")
print(f"Columns: {df.columns.tolist()}")

# Filter by category
tech_posts = df[df['category'].apply(lambda x: x.get('name', '') == 'technology')]

# Analyze engagement
print(f"Average likes: {df['stats'].apply(lambda x: x.get('likes', 0)).mean()}")

Dataset Statistics

• Total Conversations: 373
• Train Samples: 300
• Test Samples: 73
• Unique Users: 373
• Unique Categories: TBD
• Total Comments: 0
• Average Content Length: TBD characters
• Languages: Arabic (primary), English (secondary)

Use Cases

This dataset is suitable for:

• Topic Modeling: Extract main topics from Arabic conversations
• Sentiment Analysis: Analyze user sentiments and reactions
• Social Media Analysis: Study engagement patterns
• Conversational AI: Train chatbots for Arabic
• User Profiling: Analyze user behavior and interests
• Language Modeling: Build language models for Arabic
• Text Classification: Classify posts by category or topic
• Recommendation Systems: Suggest relevant content

Data Quality

• Content Completeness: All posts have full content
• Metadata Rich: Includes author, category, tags, timestamps
• Consistency: Standardized format across all samples
• Language Support: RTL (Right-to-Left) text preserved
• User Privacy: Only publicly available information

Limitations

• Comment Availability: Comments structure exists but may be empty for many posts
• Temporal Coverage: Limited to available content at scrape time
• Geographic Bias: May overrepresent certain regions
• Topic Bias: Focused on AI and intellectual topics

Ethical Considerations

• This dataset contains user-generated content
• Use responsibly and respect user privacy
• Content may reflect opinions and views of users
• Do not use for generating harmful or misleading content
• Follow the CC BY-NC-SA 4.0 license terms

Citation

If you use this dataset, please cite:

@dataset{fikran_conversations_2026,
  author = {Ouail LAAMIRI},
  title = {Fikran Conversations Dataset},
  year = {2026},
  month = {February},
  note = {Complete dataset of conversations from fikran.com, scraped in February 2026}
}

License

This dataset is licensed under Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International (CC BY-NC-SA 4.0).

Summary:

• ✅ Free to use and share
• ✅ Attribution required
• ❌ Non-commercial use only
• ✅ ShareAlike (derivative works must be under same license)

Acknowledgments

• Data source: Fikran
• Scraped using custom Python scraper
• Collected in February 2026

Contact

For questions or issues about this dataset:

• Open an issue in the repository
• Check the dataset card metadata

---

Dataset Version: 1.0.0 Last Updated: 2026-02-22