Dataset · BeIR
Nutrition scientific paper search benchmark
BeIR/nfcorpus
Corpus of English biomedical texts and queries for evaluating information retrieval systems.
What's inside
An English text dataset ranging from 1,000 to 10,000 records from the well-known BEIR benchmark. It contains queries about nutrition facts and corresponding scientific articles for the biomedical search task.
What you can do with it
- Evaluate the quality of text-retrieval models and rankers in medicine
- Test zero-shot scientific text classification systems
- Benchmark RAG pipelines on biomedical data
- Tasks
- Zero shot classification, Text retrieval
- Languages
- English
- Size
- 1K–10K rows
- Downloads this month
- 8.8K
- Likes
- 9
Models trained on it
- Stffens/bge-small-rrf-v2Embeddings 94
- racore/mega-ram-100kQuestion answering 65
- Stffens/bge-small-rrf-v3Sentence similarity 58
- nthakur/RetroMAE_BEIRSentence similarity 51
- Stffens/bge-small-rrf-v4Sentence similarity 40
- yerim0210/REPAIR_Scientific_Retrievers_0.5BEmbeddings 30
- torinriley/nfcorpus-finetuned-minilmSentence similarity 25
- MO7YW4NG/ms-marco-MiniLM-L-6-v2-4bit-nf4Text ranking 24
How to load it
from datasets import load_dataset
ds = load_dataset("BeIR/nfcorpus")