HSDSLab/FacebookMemes

---
dataset_info:
features:
- name: image
dtype: image
- name: id
dtype: string
- name: user_id
dtype: float64
- name: date
dtype: string
- name: likes
dtype: float64
- name: shares
dtype: int64
- name: comments
dtype: int64
- name: post_text
dtype: string
- name: post_link
dtype: string
- name: img_link
dtype: string
- name: ocr
dtype: string
- name: caption
dtype: string
- name: reactions
dtype: string
splits:
- name: train
num_bytes: 10804037551.84
num_examples: 235880
download_size: 8561366079
dataset_size: 10804037551.84
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
pretty_name: facebook_memes
size_categories:
- 100K<n<1M
---

# Dataset Card for Facebook Image Dataset

## Table of Contents
- [Dataset Description](#dataset-description)
- [Dataset Summary](#dataset-summary)
- [Supported Tasks and Leaderboards](#supported-tasks-and-leaderboards)
- [Languages](#languages)
- [Dataset Structure](#dataset-structure)
- [Data Instances](#data-instances)
- [Data Fields](#data-fields)
- [Data Splits](#data-splits)

## Dataset Description

### Dataset Summary
This dataset contains images scraped from Facebook along with associated metadata. The dataset is intended for research purposes, focusing on image analysis, natural language processing, and social media dynamics studies.

### Supported Tasks and Leaderboards
The dataset can be used for tasks such as image recognition, sentiment analysis, text extraction from images (OCR), and social media trend analysis.

### Languages
The text in the dataset is primarily in English, as extracted from Facebook posts.

## Dataset Structure

### Data Instances
A data instance consists of an image and its metadata. For example:
```json
{
"id": "12345",
"user_id": "67890",
"date": "2023-07-04",
"likes": 150,
"shares": 25,
"comments": 40,
"post_text": "Here's a great moment captured #fun",
"post_link": "https://facebook.com/example/status/12345",
"img_link": "https://example.com/img.jpg",
"ocr": "Here's a great moment captured",
"captions": "A boy holding a balloon.",
"reactions": ['Haha: 3 people', 'Love: 1 person']
"file_name": "12345.jpg"
}
```
### Data Fields
- `id`: Unique identifier for each post.
- `user_id`: Facebook user ID of the post author.
- `date`: Date the post was made.
- `likes`: Number of likes the post received.
- `shares`: Number of shares the post received.
- `comments`: Number of comments on the post.
- `post_text`: Text content of the post.
- `post_link`: URL to the original Facebook post.
- `img_link`: URL to the image.
- `ocr`: Text extracted from the image using OCR.
- `captions`: An image description generated by [BLIP](https://github.com/salesforce/BLIP).
- `reactions`: Reactions native to Facebook, given by users for the post.
- `file_name`: Name of the file stored locally.

### Data Splits
This dataset does not have predefined splits (train/test/validation). Users can create splits as needed for their specific tasks.