Patent Analytics Pipeline Using LLM Consensus and SBERTa Clustering

This pipeline classifies patent novelty and innovation type using a consensus of three LLMs requiring two supporting votes per label. It combines these classifications with PatentSBERTa embeddings and c-TF-IDF to map topics across large patent corpora.