Upload README.md

README.md

@@ -1,3 +1,57 @@
----
-license: apache-2.0
----
+# 📄 ddro-msmarco-document-ranking-dataset
+
+This dataset provides the **Top-300K subset** of the MS MARCO Document Ranking collection, processed for training and evaluating the DDRO (Direct Document Relevance Optimization) models.
+
+It includes:
+- Full document body text.
+- Sentence-level tokenization for improved modeling.
+- Selection based on document popularity (click counts).
+
+---
+
+### 📚 Contents
+
+| File | Description |
+|:-----|:------------|
+| `msmarco-docs-sents.top.300k.json` | Top-300K documents selected by highest click counts, with sentence tokenization. |
+
+Each entry contains:
+- `docid`
+- `url`
+- `title`
+- `body`
+- `sents` (tokenized sentences list)
+
+---
+
+### 📌 Dataset Details
+
+- **Source**:  
+  MS MARCO Document Ranking collection. [here](https://microsoft.github.io/msmarco/Datasets.html#document-ranking-dataset)
+- **Selection**:  
+  Top-300K documents ranked by click frequency from training qrels.
+- **Preprocessing**:  
+  Full body text tokenized into sentences for improved document structure.
+- **Reproducibility**:  
+  The full preprocessing pipeline is available [here](https://github.com/kidist-amde/ddro/blob/main/src/data/preprocessing/preprocess_msmarco_documents.py).
+
+
+**Note**:  
+Only the **Top-300K** split is used. The random sampling is **not used** for any experiments.
+
+---
+
+### 📖 Citation
+
+If you use this dataset, please cite:
+
+```bibtex
+@article{mekonnen2025lightweight,
+  title={Lightweight and Direct Document Relevance Optimization for Generative Information Retrieval},
+  author={Mekonnen, Kidist Amde and Tang, Yubao and de Rijke, Maarten},
+  journal={arXiv preprint arXiv:2504.05181},
+  year={2025}
+}
+```
+---
+