File size: 1,709 Bytes
602fbf7
 
dd9dd66
 
 
 
 
 
 
 
 
 
 
 
5fc762f
 
 
 
 
 
 
 
 
 
 
 
dd9dd66
602fbf7
 
 
 
 
 
 
 
 
 
 
 
 
 
dd9dd66
 
 
 
5fc762f
 
 
 
602fbf7
 
 
 
babf276
 
 
 
 
 
602fbf7
c66f6da
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
---
dataset_info:
- config_name: corpus
  features:
  - name: _id
    dtype: string
  - name: text
    dtype: string
  splits:
  - name: train
    num_bytes: 4227131
    num_examples: 2919
  download_size: 2400927
  dataset_size: 4227131
- config_name: queries
  features:
  - name: _id
    dtype: string
  - name: text
    dtype: string
  splits:
  - name: train
    num_bytes: 5352
    num_examples: 50
  download_size: 5242
  dataset_size: 5352
- config_name: relevance
  features:
  - name: query-id
    dtype: string
  - name: positive-corpus-ids
    sequence: string
  - name: bm25-ranked-ids
    sequence: string
  splits:
  - name: train
    num_bytes: 1697661
    num_examples: 50
  download_size: 244739
  dataset_size: 1697661
configs:
- config_name: corpus
  data_files:
  - split: train
    path: corpus/train-*
- config_name: queries
  data_files:
  - split: train
    path: queries/train-*
- config_name: relevance
  data_files:
  - split: train
    path: relevance/train-*
language:
- en
tags:
- sentence-transformers
size_categories:
- 1K<n<10K
---

# NanoBEIR SciFact with BM25 rankings
This dataset is an updated variant of [NanoSciFact](https://huggingface.co/datasets/zeta-alpha-ai/NanoSciFact), which is a subset of the SciFact dataset from the Benchmark for Information Retrieval (BEIR).
SciFact was created as a subset of the rather large BEIR, designed to be more efficient to run. This dataset adds a `bm25-ranked-ids` column to the `relevance` subset, which contains a ranking of every single passage in the corpus to the query.

This dataset is used in Sentence Transformers for evaluating CrossEncoder (i.e. reranker) models on NanoBEIR by reranking the top *k* results from BM25.