---
pretty_name: BEIR · NFCorpus
license: cc-by-sa-4.0
language:
- en
multilinguality:
- monolingual
task_categories:
- text-retrieval
task_ids:
- document-retrieval
tags:
- beir
- mteb
- retrieval
- biomedical
configs:
- config_name: corpus
  data_files:
  - split: test
    path: corpus/test-*.parquet
- config_name: qrels
  data_files:
  - split: test
    path: qrels/test-*.parquet
  - split: train
    path: qrels/train-*.parquet
  - split: dev
    path: qrels/dev-*.parquet
- config_name: queries
  data_files:
  - split: test
    path: queries/test-*.parquet
  - split: train
    path: queries/train-*.parquet
  - split: dev
    path: queries/dev-*.parquet
---
# NFCorpus — BEIR, unified schema

A normalised copy of the dataset behind the `mteb` task **`NFCorpus`**, one of the tasks of the [BEIR](https://github.com/beir-cellar/beir) benchmark as `mteb` defines it. Same queries, documents
and relevance judgements as the benchmark evaluates — reshaped into one strict schema shared by every dataset
in this collection.

|  |  |
|---|---|
| Source | [`mteb/nfcorpus`](https://huggingface.co/datasets/mteb/nfcorpus/tree/ec0fa4fe99da2ff19ca1214b7966684033a58814) @ `ec0fa4fe99da` (the revision pinned in `mteb`) |
| Domain · languages | biomedical · eng |
| Queries / documents / qrels (all splits) | 3,237 / 3,633 / 134,294 |
| Qrels per query | min 1 · mean 38.186 · max 475 |
| Score values | 1 ×11,758, 2 ×576 |
| Layout | `queries` · `corpus` · `qrels`, split `test`; `queries`/`qrels` also carry `train`, `dev` — one shared corpus |
| Splits | `qrels`: train, dev, test · `queries`: train, dev, test |
| License | `cc-by-sa-4.0` |

## Schema

| config | columns | rules |
|---|---|---|
| `queries` | `id: string`, `text: string` | ids unique and non-empty; every query has ≥ 1 qrel |
| `corpus` | `id: string`, `title: string`, `text: string` | `title` is always present (`""` when the source has none) |
| `qrels` | `query-id: string`, `corpus-id: string`, `score: int32` | referential integrity to both tables; no duplicate pairs; no floats |

Files are Parquet, sorted by id, zstd-compressed, sharded at 500 MB. Every rule above is enforced by a validator
before publishing; `provenance.json` records the source file hashes, what changed, and the output file hashes.

## What changed from the source

- **byte-preserved** all text — no whitespace, newline, or control-character normalisation
- `dev`: dropped 2,913 queries that have no qrels in this split (mteb does the same at load time)
- `test`: dropped 2,914 queries that have no qrels in this split (mteb does the same at load time)
- `train`: dropped 647 queries that have no qrels in this split (mteb does the same at load time)
- cast `qrels.score` double -> int32
- kept 40 documents whose text is identical to another document's — a property of the source; note that models embed such documents identically and retrieval ties among them are broken by corpus order
- renamed `_id` → `id`
- renamed source splits (`queries`←`queries/queries`, `corpus`←`corpus/corpus`) to `test`

## Load it

```python
from datasets import load_dataset
queries = load_dataset("Hyukkyu/beir-nfcorpus", "queries", split="test")
corpus  = load_dataset("Hyukkyu/beir-nfcorpus", "corpus", split="test")
qrels   = load_dataset("Hyukkyu/beir-nfcorpus", "qrels", split="test")
```

## Cross-check against the previous layout

This repo previously held the same dataset in BEIR's raw two-repo form (`Hyukkyu/beir-nfcorpus` + `Hyukkyu/beir-nfcorpus-qrels`; that layout is kept under the git tag `legacy-layout`). Before replacing it, the build was compared against it: corpus id set **identical** (3,633 documents); qrels rows per split `test` 12,334 = 12,334, `train` 110,575 = 110,575, `dev` 11,385 = 11,385.

## License and attribution

The data is redistributed under the source's terms — `cc-by-sa-4.0`. All credit belongs to the
original authors; see the source repository and the references in `mteb`'s task metadata (https://www.cl.uni-heidelberg.de/statnlpgroup/nfcorpus/).
This repository is an independent repackaging and is not affiliated with the RTEB or MTEB maintainers.

> **License discrepancy.** The upstream dataset card declares `cc-by-sa-4.0` while `mteb`'s `TaskMetadata` declares `not specified`. This card carries the upstream value as the more conservative choice.