README.md

---
license: cc0-1.0
task_categories:
- question-answering
- feature-extraction
language:
- en
pretty_name: PaperSeek OpenAlex Embeddings
size_categories:
- 100M<n<1B
---
# 📚 PaperSeek: OpenAlex English Titles & Abstracts (April 2025 Snapshot)

This dataset is part of the **[PaperSeek](https://github.com/Mohammadsaknini/PaperSeek)** framework, a semantic search engine designed for literature discovery using research questions and prior knowledge. PaperSeek is developed as part of a Master's thesis to explore novel approaches in enhancing academic search relevance.

## 📦 Dataset Overview

- **Source**: [OpenAlex](https://openalex.org/)
- **Snapshot Date**: April 1st, 2025
- **Language**: English
- **Contents**:
  - Title
  - Abstract
  - Embedding

The dataset includes **102 million** research works, with English-language titles and abstracts extracted from the OpenAlex snapshot published on April 1st, 2025.

## 🧠 Embeddings

- **Model**: [Stella V5 1.5B]([https://huggingface.co/) (bfloat16 precision)
- **Embedding Prompt**: Title: {title}\n[SEP] Abstract: {abstract}

- **Output**: Dense embeddings suitable for semantic search and large-scale information retrieval.

## 🧪 Use Cases

- Semantic search
- Embedding-based retrieval
- Academic literature analysis
- NLP benchmarking on large-scale scientific text

## 📝 Citation

If you use this dataset, please cite/star the [PaperSeek](https://github.com/Mohammadsaknini/PaperSeek) project and OpenAlex as data sources. A formal citation for the Master's thesis will be provided once available.

## 📁 Data Structure

Each entry contains:
- `id`: Unique work ID (OpenAlex identifier)
- `title`: English title of the research work
- `abstract`: English abstract of the research work
- `embedding`: Float32 ncoded vector

## ⚠️ License & Use

- OpenAlex data is released under the [CC0 1.0 Universal (Public Domain Dedication)](https://creativecommons.org/publicdomain/zero/1.0/).
- Embeddings generated as part of the PaperSeek framework are released for research purposes.

## 🙋‍♂️ Contact

For questions or collaborations, feel free to get in touch through Hugging Face or via the Github.