FlavioSoriano/wos-46985-rag-fuse

# WOS-46985 for RAG-Fuse

This repository snapshot packages the raw WOS-46985 dataset files used during the RAG-Fuse onboarding work, so they can be downloaded easily from any server through the Hugging Face Hub.

## Contents

- `raw/WOS46985/X.txt`
- `raw/WOS46985/Y.txt`
- `raw/WOS46985/YL1.txt`
- `raw/WOS46985/YL2.txt`
- `raw/WOS46985/Meta-data/Data.xlsx`

## Source

- Canonical dataset page: https://data.mendeley.com/datasets/9rw3vkcfy4/2
- DOI: `10.17632/9rw3vkcfy4.2`
- License: `CC BY 4.0`

## Semantics

- `Y` is the canonical flat label id with 134 classes.
- `YL1` is the parent/domain label with 7 classes.
- `YL2` is a level-2 code reused across parents and must not be treated as a globally unique flat label id.
- `Meta-data/Data.xlsx` is the official metadata sheet used to map `Y -> area` when preparing label names for RAG-Fuse.

## Intended use in this project

This package is meant to simplify transfer and download of the raw WOS-46985 files across machines.
The RAG-Fuse preparation flow should keep `Y` as the target id and use the metadata only to name labels.