gmp007/PropertyExtractor

Combines zero-shot and few-shot in-context learning with engineered dynamic prompts to extract material properties from unstructured scientific papers, achieving >90% precision and recall. Supports both Google Gemini Pro and OpenAI GPT-4 backends with configurable LLM parameters (temperature, token limits). Integrates with academic APIs (ScienceDirect, CrossRef, PubMed) for sourcing unstructured data and outputs structured material property databases via CSV/Excel workflows.