File size: 2,024 Bytes
0850e9a
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
---
license: apache-2.0
---
# PROSPERO Inclusion/Exclusion Criteria Dataset

This dataset is a curated and preprocessed collection of clinical research objectives and their corresponding inclusion and exclusion criteria, extracted from the [PROSPERO](https://www.crd.york.ac.uk/prospero/) international prospective register of systematic reviews.

## Description

The dataset was constructed to support fine-tuning of large language models (LLMs) for tasks involving the automated generation of eligibility criteria based on research objectives. It contains English-language research objectives from systematic review protocols across three domains:

- Mental health
- Diagnostics
- Therapeutics

## Dataset Structure

Each row in the dataset corresponds to one study protocol and contains:

- `Study_ID`: The unique identifier from PROSPERO (e.g., CRD420250616527)
- `Objectives`: The research objective(s) of the systematic review
- `Included`: The criteria defining which participants/studies are eligible for inclusion
- `Excluded`: The criteria used to disqualify participants/studies from inclusion


## Use Case

This dataset is ideal for training or fine-tuning LLMs (like T5, Mistral, LLaMA) on:

- Generating inclusion/exclusion criteria from research objectives  
- Biomedical text-to-text reasoning tasks  
- Automating systematic review assistance and meta-research workflows

## Format

- File type: CSV (`utf-8`)
- Fields: `Study_ID`, `Objectives`, `Included`, `Excluded`
- Suggested fields for fine-tuning:
  - `input_text`: Prompt generated from `Objectives`
  - `output_text`: "Inclusion: ... Exclusion: ..." combined string

## License & Attribution

Data is sourced from PROSPERO, a publicly accessible database, and is intended for educational and research purposes only. Please cite the original PROSPERO records when using this dataset in any published work.

## 🙌 Created By

Sanaa as part of her mission to automate and enhance systematic literature reviews using AI.
email: sanaa.abril@gmail.com