pyterrier/scifact.terrier

---
# 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
---

# scifact.terrier

## Description

Terrier index for SciFact

## Usage

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

## Benchmarks

| name | nDCG@10 | R@1000 |
|:-------|----------:|---------:|
| bm25 | 0.6839 | 0.9767 |
| dph | 0.6716 | 0.9733 |

## Reproduction

```python
import pyterrier as pt
from tqdm import tqdm
import ir_datasets
dataset = ir_datasets.load('beir/scifact')
meta_docno_len = dataset.metadata()['docs']['fields']['doc_id']['max_len']
indexer = pt.IterDictIndexer("./scifact.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"
}
```

# pretty_name: "" # 示例："MS MARCO Terrier 索引"
tags:
- PyTerrier (PyTerrier)
- pyterrier-artifact
- pyterrier-artifact.sparse_index
- pyterrier-artifact.sparse_index.terrier
task_categories:
- 文本检索
可视化查看器: 禁用
---

# scifact.terrier

## 描述

本数据集为 SciFact 对应的 Terrier 稀疏索引

## 使用示例

python
# 加载该制品
import pyterrier as pt
index = pt.Artifact.from_hf('pyterrier/scifact.terrier')
index.bm25()

## 基准测试指标

| 模型名称 | nDCG@10 | 召回率@1000 |
|:-------|----------:|---------:|
| bm25 | 0.6839 | 0.9767 |
| dph | 0.6716 | 0.9733 |

## 复现流程

python
import pyterrier as pt
from tqdm import tqdm
import ir_datasets
dataset = ir_datasets.load('beir/scifact')
meta_docno_len = dataset.metadata()['docs']['fields']['doc_id']['max_len']
indexer = pt.IterDictIndexer("./scifact.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)

## 元数据

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