PRISMA Search Literature Review Tool

Overview

Automate and document literature reviews using the PRISMA methodology. Input a research topic or keywords, manage PRISMA data, and output publication lists and methodology text for research papers.

Disclaimer:
Review and verify results before including them in research. Automated outputs assist, not replace, expert judgment.
API Usage Notice:
Use APIs responsibly to avoid rate limiting or loss of access.

Features

• Flexible keyword input: list or comma-separated string
• Smart keyword handling: derives keywords from research topic if not provided
• Enhanced output: multiple CSVs for all and selected publications
• Collects PRISMA-related values (inclusion/exclusion criteria, databases, date ranges)
• NEW: Database-specific field mapping for inclusion/exclusion criteria (e.g., type:journal, language:english)
• Outputs:
  • PRISMA method data for literature review
  • CSVs: all_publications_found.csv, selected_publications.csv, output_results.csv
  • JSON file with results
  • PRISMA flow diagram in draw.io format (default or user template)
• Supports multiple databases: PubMed, CrossRef, arXiv, CORE, SemanticScholar, IEEE Xplore, Springer, DBLP, Scopus
• Validates configuration files
• Retries API requests with exponential backoff
• Modular, extensible design

PRISMA Flow Diagram Template

Default template:
src/literature_search/prisma_flow_diagram.drawio
User template: place prisma_flow_diagram.drawio in output directory.
Filled diagram: prisma_flow_diagram_filled.drawio in output directory.

Configuration

Keywords

Provide keywords as a list or comma-separated string:

{
  "keywords": ["machine learning", "deep learning", "neural networks"]
}

or

{
  "keywords": "machine learning, deep learning, neural networks"
}

If missing, keywords are generated from research_topic with a warning.

Sample Configuration

{
  "research_topic": "fundus image dataset",
  "keywords": "medical-imaging, machine-learning, fundus, retinal, ophthalmology",
  "initial_prisma_values": {
    "inclusion_criteria": ["review", "thesis", "journal", "book"],
    "exclusion_criteria": ["non-english", "conference"],
    "databases": ["PubMed", "CrossRef", "arXiv", "CORE", "SemanticScholar", "IEEE", "Springer", "DBLP", "Scopus"],
    "date_range": "2015-2025"
  },
  "api_keys": {
    "CORE": "your_core_api_key_here",
    "IEEE": "your_ieee_api_key_here", 
    "Springer": "your_springer_api_key_here",
    "Scopus": "your_scopus_api_key_here"
  }
}

API Keys Configuration

Some databases require API keys for access:

• CORE: Free API key available at CORE API
• IEEE Xplore: API key required from IEEE Developer
• Springer: API key required from Springer Nature API
• Scopus: API key required from Elsevier Developer Portal

Add your API keys to the configuration file under the api_keys section. If no API key is provided for a database that requires one, that database will be skipped with a warning message.

Field-Specific Criteria (NEW)

You can now specify database-specific fields in inclusion and exclusion criteria using the format field:value:

{
  "research_topic": "machine learning",
  "initial_prisma_values": {
    "inclusion_criteria": [
      "type:journal-article",
      "language:english",
      "journal:nature"
    ],
    "exclusion_criteria": [
      "type:conference-paper",
      "source:arxiv",
      "language:non-english"
    ],
    "databases": ["PubMed", "CrossRef", "arXiv"],
    "date_range": "2020-2025"
  }
}

Supported fields:

• type / publication_type / pubtype → Publication type (journal, conference, etc.)
• language → Publication language
• source → Database source
• journal / venue → Journal or venue name
• authors → Author names
• document_type → Document type (CORE database)

Database-specific mappings:

• PubMed/Europe PMC: Maps to publication metadata from medical literature
• CrossRef: Maps to scholarly publication metadata
• arXiv: Maps to preprint metadata
• CORE: Maps document_type to Type field
• SemanticScholar: Maps venue to Journal field

Backward compatibility: Criteria without field specifications (e.g., "journal") default to the type field.

Installation

Install via pip:

pip install literature_search

Or from wheel:

pip install dist/literature_search-*.whl

Usage

Console Script

Run from anywhere:

literature-search --config sample_input.json --logic OR --page_size 100 --output_dir output

Arguments:

• --config: Path to config JSON (default: sample_input.json)
• --logic: Keyword logic ('AND' or 'OR', default: OR)
• --page_size: Results per database (default: 100)
• --output_dir: Output directory (default: output)

Programmatic Usage

Import and use in Python:

from literature_search.config_loader import load_config
from literature_search.keywords import get_keywords
from literature_search.api_clients import (
    get_publications_europe_pmc,
    get_publications_crossref,
    get_publications_arxiv,
    get_publications_core,
    get_publications_semanticscholar,
    get_publications_ieee,
    get_publications_springer,
    get_publications_dblp,
    get_publications_scopus
)
from literature_search.prisma_logs import output_prisma_results, create_prisma_drawio_diagram

config = load_config('sample_input.json')
keywords = get_keywords(config.get('research_topic', ''))
publications = get_publications_europe_pmc(keywords)
output_prisma_results(publications, criteria_counts, total_records, output_dir='output')
create_prisma_drawio_diagram(criteria_counts, total_records, output_dir='output')

Output Files

• all_publications_found.csv: All publications with inclusion/exclusion info
• selected_publications.csv: Publications meeting inclusion criteria
• output_results.csv: All publications with inclusion status
• results.json: Complete results
• prisma_flow_diagram.drawio: PRISMA flow diagram

Requirements

• Ubuntu 24.04.2 LTS
• Python 3.10+
• Python packages: requests, drawpyo, matplotlib, pytest

Contributing

See .github/copilot-instructions.md for coding standards.

Contribution Guide

1. Fork and create a feature branch.
2. Follow PEP 8 and PRISMA methodology.
3. Write clear commit messages and documentation.
4. Add tests for new features or fixes.
5. Submit a pull request.

Package Publication

The package is automatically published to PyPI when:

1. Changes are merged to the main branch, AND
2. The commit message starts with [release]

For example: [release] Version 1.0.0 with new features

The publication uses PyPI OIDC (OpenID Connect) for secure authentication without requiring API tokens.

License

GPL-3

Future Work

• Add more database APIs
• AI-based abstract inclusion/exclusion