README.md

---
license: cc-by-nc-4.0
task_categories:
- image-to-text
language:
- tr
tags:
- image
- captioning
- image captioning
- turkish
- e-commerce
- e-ticaret
pretty_name: Turkish E-Commerce Product Captions
size_categories:
- 100K<n<1M
extra_gated_prompt: >
  This dataset requires manual approval due to licensing and ethical considerations.
  Please briefly describe your intended use and include any relevant academic affiliation or reasearch context.
dataset_info:
  features:
  - name: image
    dtype: image
  - name: text
    dtype: string
  splits:
  - name: train
    num_bytes: 24409358046.884132
    num_examples: 168516
  - name: validation
    num_bytes: 3037953072.282933
    num_examples: 21065
  - name: test
    num_bytes: 3034286730.0979333
    num_examples: 21065
  download_size: 29341642373
  dataset_size: 30481597849.265
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
  - split: validation
    path: data/validation-*
  - split: test
    path: data/test-*
---
# Dataset Card for Turkish E-Commerce Product Captions

## Dataset Details

### Dataset Description

This dataset consists of Turkish captions for product images collected from publicly accessible pages on Trendyol. The captions were generated using a captioning model and curated manually by the dataset author for academic and research purposes. The dataset includes approximately 200K samples and is intended for image captioning tasks in Turkish.

- **Curated by:** Mustafa Batuhan Değirmenci
- **Funded by:** N/A
- **Shared by:** Mustafa Batuhan Değirmenci
- **Language(s):** Turkish
- **License:** CC BY-NC 4.0

### Dataset Sources

- **Images Source:** Publicly available product images from [Trendyol](https://www.trendyol.com)
- **Captions:** Generated and edited by the author for research

## Uses

### Direct Use

- Academic research in multimodal learning (image-to-text)
- Training and evaluation of Turkish image captioning models
- Benchmarking captioning models on e-commerce content

### Out-of-Scope Use

- Commercial applications are prohibited due to copyright restrictions.
- Redistribution of original images is not allowed without permission from Trendyol or sellers.

## Access
This dataset is gated. You need to **request access** to download it.
Access will be granted for **non-commercial academic use** only.
Please briefly explain your research purpose when submitting your access request.

## Dataset Structure

Each example contains:
- `image`: a product image in RGB format
- `text`: a Turkish caption describing the image content

No additional metadata such as product category or price is included.

## Dataset Creation

### Curation Rationale

The dataset was created to address the lack of publicly available Turkish image-captioning datasets, especially in the e-commerce domain.

### Source Data

#### Data Collection and Processing

- Images were collected from public Trendyol product listings using a scraper.
- Captions were generated using an image captioning model and curated by the author.
- All captions were written in Turkish.

#### Who are the source data producers?

- Images were created by Trendyol vendors.
- Captions were generated by the dataset author using a trained model.

### Annotations

#### Annotation process

- Model-generated captions were reviewed and filtered for quality.
- No external annotators were involved.

#### Who are the annotators?

- The author (Mustafa Batuhan Değirmenci)

#### Personal and Sensitive Information

No personal or sensitive information is included in the dataset.

## Bias, Risks, and Limitations

- Captions are generated by a model and may contain inaccuracies.
- Dataset reflects products and styles from a specific e-commerce platform (Trendyol), introducing domain-specific bias.

### Recommendations

- Avoid deploying models trained on this dataset in sensitive or commercial settings without additional validation.

## Citation

If you use this dataset, please cite:
```bibtex
@misc{turkish-ecommerce-captions,
  author = {Mustafa Batuhan Değirmenci},
  title = {Turkish E-Commerce Product Captions},
  year = {2025},
  note = {Dataset available on Hugging Face Hub}
}
```


## Contact

For questions or feedback: [batuhan.degirmenci@outlook.com]