ReviewAid

ReviewAid automates full-text screening and structured data extraction in systematic reviews by using large language models to classify articles according to user-defined PICO criteria and extract information from full-text publications.

Key Features:

• Multi-provider AI architecture: Supports multiple large language model providers including OpenAI, Anthropic, DeepSeek, Cohere, Z.ai, and local execution via Ollama.
• PICO-based screening: Classifies articles according to Population, Intervention, Comparison, and Outcome criteria for structured literature identification in healthcare evidence synthesis.
• Customizable data extraction: Extracts structured information from full-text publications with user-configurable parameters.
• Bulletproof Parsing Pipeline: Recovers data from malformed AI outputs such as text with broken formatting to prevent software crashes.
• Hierarchical confidence scoring: Employs a four-tier system to quantify the certainty of AI classification and extraction decisions.

Scientific Applications:

• Systematic review automation: Accelerates evidence synthesis workflows by reducing manual screening workload while maintaining human oversight as a supplementary reviewer.
• Healthcare literature screening: Identifies relevant studies from large publication sets using structured PICO criteria for clinical and epidemiological research.
• Structured data extraction: Retrieves organized information from full-text articles to support meta-analysis and evidence-based research.

Methodology:

ReviewAid uses large language models for PICO-based article classification and structured data extraction from full-text publications, with a Bulletproof Parsing Pipeline to handle malformed outputs and a four-tier confidence scoring system for decision quantification.