File size: 1,793 Bytes
c727fd7
 
 
 
 
 
b620107
 
 
 
c727fd7
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
# FactStatement Dataset

## Overview

FactStatement is the foundational layer of FactNet, a cross-lingual, multi-layered fact knowledge graph. FactStatements are language-neutral, atomic fact units directly mapped from Wikidata statements, forming the core building blocks of the knowledge graph.

+ Paper: https://arxiv.org/abs/2602.03417
+ Github: https://github.com/yl-shen/factnet
+ Dataset: https://huggingface.co/collections/openbmb/factnet

## Dataset Format

The dataset contains parquet files with the following key fields:

- `core_id`: Unique identifier for the fact statement
- `subject_qid`: Wikidata QID of the subject entity
- `property_pid`: Wikidata PID of the property
- `value`: Raw value from Wikidata
- `value_qid`: Entity QID if the value is an entity (otherwise null)
- `normalized_value`: Standardized representation of the value
- `claim_hash`: Standardized hash of subject|property|normalized_value|qualifiers
- `qualifiers`: Qualifier information (time, location, etc.)
- `references`: Source information
- `rank`: Wikidata rank (preferred, normal, deprecated)
- `confidence`: Computed confidence score

## Usage

FactStatements are designed to be language-neutral representations of facts that can be linguistically realized through the FactSense layer and semantically grouped through the FactSynset layer.

## License

This dataset is derived from Wikidata and is available under the CC0 license.

## Citation

```
@article{shen2026factnet,
  title={FactNet: A Billion-Scale Knowledge Graph for Multilingual Factual Grounding},
  author={Shen, Yingli and Lai, Wen and Zhou, Jie and Zhang, Xueren and Wang, Yudong and Luo, Kangyang and Wang, Shuo and Gao, Ge and Fraser, Alexander and Sun, Maosong},
  journal={arXiv preprint arXiv:2602.03417},
  year={2026}
}
```