This paper describes a reproducible pipeline for target normalization and knowledge graph (KG) linking that integrates with the structured syntactic–semantic alignment and calibration primitives developed in preceding program papers (P1–P4). The pipeline combines a scalable FAISS-based dense retrieval stage, a context coherence reranker that fuses textual, document-level, and KG structural signals, and a structured calibration layer that produces well-calibrated link confidences suitable for downstream decision thresholds and human-in-the-loop review. The manuscript documents algorithmic details, a formal calibration bound, provenance and adversarial evaluation protocols (explicitly attributing borrowed methods), and a completesupplementary bundle (proofs, annotation protocol, reproducibility scripts, and ASCII figures) to support reproducibility and ethical attribution
Usman Zafar (Tue,) studied this question.