Update README.md

README.md

@@ -6,4 +6,37 @@ language:
 - en
 size_categories:
 - 10M<n<100M
----
+---
+
+
+![Descriptor logo](DocDescriptorLogo_smaller.png)
+
+Data release for the paper Task-Agnostic Web Document Annotation with LLM-Generated Descriptors (forthcoming).
+
+This dataset contains a descriptor-annotated version of a 10 billion token (~14 million document) [sample of FineWeb](https://huggingface.co/datasets/HuggingFaceFW/fineweb/viewer/sample-10BT). The descriptors are generated via a task-agnostic data annotation pipeline described in the paper (link coming soon).
+
+### Data fields
+- text: Document text
+- doc_id: Document ID (these match those in the original FineWeb release)
+- harmonized_descriptors: Schema mapped descriptors
+- raw_descriptors: Free-form descriptors that do not follow the schema
+
+### Citing
+If you use this data please cite the original FineWeb paper and the our paper.
+
+**Task-Agnostic Web Document Annotation with LLM-Generated Descriptors**
+```text
+Coming soon.
+```
+
+**FineWeb**: Link to ArXiv: https://arxiv.org/abs/2406.17557
+```text
+@inproceedings{
+penedo2024the,
+title={The FineWeb Datasets: Decanting the Web for the Finest Text Data at Scale},
+author={Guilherme Penedo and Hynek Kydl{\'\i}{\v{c}}ek and Loubna Ben allal and Anton Lozhkov and Margaret Mitchell and Colin Raffel and Leandro Von Werra and Thomas Wolf},
+booktitle={The Thirty-eight Conference on Neural Information Processing Systems Datasets and Benchmarks Track},
+year={2024},
+url={https://openreview.net/forum?id=n6SCkn2QaG}
+}
+```