README.md

---
# pretty_name: "" # Example: "MS MARCO Terrier Index"
tags:
- pyterrier
- pyterrier-artifact
- pyterrier-artifact.sparse_index
- pyterrier-artifact.sparse_index.terrier
task_categories:
- text-retrieval
viewer: false
---

# nfcorpus.terrier

## Description

Terrier index for NFCorpus

## Usage

```python
# Load the artifact
import pyterrier as pt
index = pt.Artifact.from_hf('pyterrier/nfcorpus.terrier')
index.bm25()
```

## Benchmarks

`nfcorpus/dev`

| name   |   nDCG@10 |   R@1000 |
|:-------|----------:|---------:|
| bm25   |    0.2944 |   0.3291 |
| dph    |    0.2914 |   0.3293 |

`nfcorpus/test`

| name   |   nDCG@10 |   R@1000 |
|:-------|----------:|---------:|
| bm25   |    0.3282 |   0.365  |
| dph    |    0.3221 |   0.3671 |

## Reproduction

```python
import pyterrier as pt
from tqdm import tqdm
import ir_datasets
dataset = ir_datasets.load('beir/nfcorpus')
meta_docno_len = dataset.metadata()['docs']['fields']['doc_id']['max_len']
indexer = pt.IterDictIndexer("./nfcorpus.terrier", meta={'docno': meta_docno_len, 'text': 4096})
docs = ({'docno': d.doc_id, 'text': d.default_text()} for d in tqdm(dataset.docs))
indexer.index(docs)
```

## Metadata

```
{
  "type": "sparse_index",
  "format": "terrier",
  "package_hint": "python-terrier"
}
```