File size: 2,615 Bytes
0ec1013
eaa31e4
 
420e5eb
eaa31e4
 
0ec1013
 
 
 
 
 
 
 
 
 
 
 
f5b3324
 
0ec1013
f5b3324
0ec1013
 
 
f5b3324
 
0ec1013
f5b3324
 
0ec1013
 
 
 
 
 
 
eaa31e4
0ec1013
 
 
 
202e832
 
 
 
 
 
 
 
 
0ec1013
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
---
language:
- en
license: cc-by-sa-3.0
size_categories:
- 1K<n<10K
task_categories:
- summarization
tags:
- biomedical
- health
- NLP
- summarization
- LLM
- factuality
---

PlainFact is a high-quality human-annotated dataset with fine-grained explanation (i.e., added information) annotations designed for Plain Language Summarization tasks, along with [PlainQAFact](https://github.com/zhiwenyou103/PlainQAFact) factuality evaluation framework. It is collected from the [Cochrane database](https://www.cochranelibrary.com/) sampled from CELLS dataset ([Guo et al., 2024](https://doi.org/10.1016/j.jbi.2023.104580)). 
PlainFact is a sentence-level benchmark that splits the summaries into sentences with fine-grained explanation annotations. In total, we have 200 plain language summary-abstract pairs (2,740 sentences).
In addition to all factual plain language sentences, we also generate contrasting non-factual examples for each plain language sentence. These contrasting examples are perturbed using GPT-4o, following the perturbation criteria for faithfulness introduced in APPLS ([Guo et al., 2024](https://aclanthology.org/2024.emnlp-main.519/)).

> Currently, we only released the annotation for **Explanation** sentences. We will release the full version of PlainFact soon (including Category and Relation information). Stay tuned!


Here are explanations for the headings:
- **Target_Sentence_factual**: The all factual plain language sentence.
- **Target_Sentence_non_factual**: The perturbed (non-factual) plain language sentence.
- **External**: Whether the sentence includes information does not explicitly present in the scientific abstract. (yes: explanation, no: simplification)
- **Original_Abstract**: The scientific abstract corresponding to each sentence/summary.


You can load our dataset as follows:
```python
from datasets import load_dataset
plainfact = load_dataset("uzw/PlainFact")
```

For detailed information regarding the dataset or factuality evaluation framework, please refer to our [Github repo](https://github.com/zhiwenyou103/PlainQAFact) and paper at https://huggingface.co/papers/2503.08890.

Citation
If you use data from PlainFact or PlainFact-summary, please cite with the following BibTex entry:
```
@misc{you2025plainqafactautomaticfactualityevaluation,
      title={PlainQAFact: Automatic Factuality Evaluation Metric for Biomedical Plain Language Summaries Generation}, 
      author={Zhiwen You and Yue Guo},
      year={2025},
      eprint={2503.08890},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2503.08890}, 
}
```