Upload README.md with huggingface_hub

README.md

@@ -1,120 +1,59 @@
 ---
 license: mit
 task_categories:
-- token-classification
+- named-entity-recognition
+- named-entity-linking
+- text-classification
 language:
 - en
 tags:
 - biology
 - proteins
 - genes
-- named-entity-linking
 - uniprot
-- sourcedata
-- nel
+- named-entity-linking
+- biomedical
+- life-sciences
 size_categories:
 - 10K<n<100K
-configs:
-- config_name: default
-  data_files:
-  - split: train
-    path: data/train-*
-  - split: validation
-    path: data/validation-*
-  - split: test
-    path: data/test-*
-dataset_info:
-  features:
-  - name: pmid
-    dtype: string
-  - name: doi
-    dtype: string
-  - name: title
-    dtype: string
-  - name: abstract
-    dtype: string
-  - name: year
-    dtype: string
-  - name: fig_label
-    dtype: string
-  - name: caption
-    dtype: string
-  - name: fig_href
-    dtype: string
-  - name: entities
-    list:
-    - name: ext_dbs
-      dtype: string
-    - name: ext_ids
-      dtype: string
-    - name: ext_tax_ids
-      dtype: string
-    - name: ext_tax_names
-      dtype: string
-    - name: mapping_source
-      dtype: string
-    - name: mapping_status
-      dtype: string
-    - name: ncbi_gene_id
-      dtype: string
-    - name: original_type
-      dtype: string
-    - name: role
-      dtype: string
-    - name: text
-      dtype: string
-    - name: type
-      dtype: string
-    - name: uniprot_ids
-      list: string
-  splits:
-  - name: train
-    num_bytes: 63526370
-    num_examples: 14880
-  - name: validation
-    num_bytes: 3301935
-    num_examples: 825
-  - name: test
-    num_bytes: 3293603
-    num_examples: 827
-  download_size: 17320950
-  dataset_size: 70121908
 ---
 
-# SODA-NEL: SourceData Named Entity Linking Dataset
+# SODA-SPROUT: Role-Filtered Named Entity Linking Dataset
 
 ## Dataset Description
 
-SODA-NEL is a comprehensive Named Entity Linking (NEL) dataset extracted from the [SourceData database](https://huggingface.co/datasets/EMBO/SourceData), containing 16,889 figure-level records with 194,383 mapped protein/gene entities. This dataset is specifically designed for training and evaluating Named Entity Recognition (NER) and Named Entity Linking (NEL) models in the life sciences domain, with a focus on protein/gene normalization to UniProt accession numbers.
+This dataset is an improved version of the SODA-SPROUT NEL dataset, specifically filtered to include only the most relevant biological entities for Named Entity Linking tasks. The dataset focuses on proteins and genes with roles of **'assayed'** and **'intervention'**, which represent the core biological entities that are actually measured or manipulated in scientific experiments.
 
-## Dataset Statistics
+### Key Improvements
 
-- **Total Records**: 16,889 figure-level records
-- **Train**: 15,205 records (90.0%)
-- **Validation**: 812 records (4.8%)
-- **Test**: 872 records (5.2%)
-- **Entities**: 194,383 successfully mapped to UniProt (92.2% mapping success)
-- **NCBI Gene Mapping**: 7,942 out of 8,489 successfully mapped (93.6%)
+- **Role Filtering**: Only includes entities with roles 'assayed' (104,286 entities) and 'intervention' (80,127 entities)
+- **Higher Quality**: Removes less relevant entities like 'reporter', 'component', 'normalizing', and 'experiment' roles
+- **Better NEL Performance**: Focuses on the most important biological entities for linking tasks
+- **Reduced Noise**: 16,532 figure-level records (vs. 21,000+ in unfiltered version)
 
 ## Dataset Structure
 
-Each record contains:
+### Statistics
+- **Total Records**: 16,532 figure-level records
+- **Training Set**: 14,880 records (90%)
+- **Validation Set**: 825 records (5%)
+- **Test Set**: 827 records (5%)
+- **Total Entities**: 184,413 (104,286 assayed + 80,127 intervention)
+- **Unique Papers**: 3,101 papers
+
+### Data Format
 
-- **paper_id**: SourceData paper identifier
-- **figure_id**: SourceData figure identifier
-- **caption**: Figure caption text
-- **abstract**: Paper abstract
-- **organisms**: List of organism names
-- **entities**: List of protein/gene entities with:
+Each record contains:
+- **Paper Context**: PMID, DOI, title, abstract, year
+- **Figure Context**: Figure label, caption, figure URL
+- **Entities**: List of gene/protein entities with:
   - `text`: Entity mention text
-  - `start`: Start position in caption
-  - `end`: End position in caption
-  - `type`: Entity type (always "geneprod" for NER compatibility)
-  - `original_type`: Original SourceData type ("gene" or "protein")
-  - `ncbi_gene_id`: NCBI Gene ID (if available)
-  - `uniprot_ids`: List of UniProt accession numbers
+  - `type`: "geneprod" (for NER compatibility)
+  - `original_type`: "gene" or "protein" (from SourceData)
+  - `role`: "assayed" or "intervention"
+  - `ext_tax_names`: Organism taxonomy name
+  - `uniprot_ids`: Ground truth UniProt accessions
   - `mapping_status`: "mapped" or "unmapped"
-  - `mapping_source`: Source of mapping ("uniprot_direct" or "ncbi_gene_mapping")
 
 ## Usage
 
@@ -124,77 +63,49 @@ from datasets import load_dataset
 # Load the dataset
 dataset = load_dataset("EMBO/soda-nel")
 
-# Access train/validation/test splits
+# Access splits
 train_data = dataset["train"]
-val_data = dataset["validation"]
+val_data = dataset["validation"] 
 test_data = dataset["test"]
 
 # Example record
 record = train_data[0]
+print(f"Paper: {record['title']}")
+print(f"Figure: {record['fig_label']}")
 print(f"Caption: {record['caption']}")
-print(f"Entities: {record['entities']}")
+print(f"Entities: {len(record['entities'])}")
 ```
 
-## Data Sources
-
-This dataset is derived from the [SourceData dataset](https://huggingface.co/datasets/EMBO/SourceData), which contains 60k+ life sciences papers with expert-curated annotations. SODA-NEL specifically focuses on:
-
-- **Source**: [SourceData database](https://huggingface.co/datasets/EMBO/SourceData) (60k+ life sciences papers)
-- **Ground Truth**: UniProt accession numbers
-- **Entity Types**: Gene and protein mentions from figure captions
-- **Organisms**: Multiple organisms per paper/entity
-- **Filtering**: Only figures with gene/protein entities (`in_caption=True`)
+## Role Definitions
 
-## Related Work
+- **Assayed**: Proteins/genes that were actually measured, quantified, or tested in the experiment
+- **Intervention**: Proteins/genes that were manipulated, targeted, or used as experimental interventions
 
-This dataset builds upon the foundational work of the SourceData project:
+## Data Sources
 
-- **Original SourceData Dataset**: [EMBO/SourceData](https://huggingface.co/datasets/EMBO/SourceData)
-- **SourceData Paper**: [The SourceData-NLP dataset: integrating curation into scientific publishing for training large language models](https://arxiv.org/abs/2310.20440)
-- **SourceData Platform**: [sourcedata.embo.org](https://sourcedata.embo.org)
+- **Source Dataset**: [EMBO/SourceData](https://huggingface.co/datasets/EMBO/SourceData)
+- **Original Paper**: [SourceData: A platform for the large-scale curation of unstructured data from the scientific literature](https://arxiv.org/abs/2310.20440)
+- **UniProt Mapping**: NCBI Gene IDs mapped to UniProt accessions via UniProt REST API
 
 ## Citation
 
-If you use this dataset, please cite both the original SourceData work and this derived dataset:
-
-```bibtex
-@dataset{soda_nel_2025,
-  title={SODA-NEL: SourceData Named Entity Linking Dataset},
-  author={EMBO},
-  year={2025},
-  publisher={HuggingFace},
-  url={https://huggingface.co/datasets/EMBO/soda-nel}
-}
-```
-
-And the original SourceData dataset:
+If you use this dataset, please cite both the original SourceData paper and this improved version:
 
 ```bibtex
-@ARTICLE{2023arXiv231020440A,
-       author = {{Abreu-Vicente}, Jorge and {Sonntag}, Hannah and {Eidens}, Thomas and {Lemberger}, Thomas},
-        title = "{The SourceData-NLP dataset: integrating curation into scientific publishing for training large language models}",
-      journal = {arXiv e-prints},
-     keywords = {Computer Science - Computation and Language},
-         year = 2023,
-        month = oct,
-          eid = {arXiv:2310.20440},
-        pages = {arXiv:2310.20440},
-archivePrefix = {arXiv},
-       eprint = {2310.20440},
- primaryClass = {cs.CL},
-       adsurl = {https://ui.adsabs.harvard.edu/abs/2023arXiv231020440A},
-      adsnote = {Provided by the SAO/NASA Astrophysics Data System}
+@article{source_data_2023,
+  title={SourceData: A platform for the large-scale curation of unstructured data from the scientific literature},
+  author={...},
+  journal={...},
+  year={2023},
+  url={https://arxiv.org/abs/2310.20440}
 }
 ```
 
 ## License
 
-This dataset is released under the MIT License, following the same license as the original SourceData dataset.
-
-## Contact
-
-For questions or issues, please contact the EMBO team or open an issue on the repository.
+This dataset is released under the MIT License. The underlying SourceData is available under its original license.
 
-## Acknowledgments
+## Updates
 
-We thank the SourceData team for creating the foundational dataset and the EMBO organization for supporting this work.
+- **v1.1.0**: Added role filtering for 'assayed' and 'intervention' entities only
+- **v1.0.0**: Initial release with all entity roles