Reinforcement Learning for Multi-Project Scheduling and Resource Allocation (RL-MPSRA)

Abstract

Semantic enrichment research has diversified from ontology-centric annotation to transformer-era methods for entity/relation induction and alignment. Building on the bibliometric baseline in [ 1], we present a longitudinal analysis of topic drift and cross-topic influence between 2008 and 2024. We couple a dynamic topic model with a citation-graph Hawkes process to quantify (i) drift rates of macro-themes and (ii) excitation pathways by which influential works shift attention across themes. On a curated corpus of 5,412 Scopus-indexed records, our model improves held-out perplexity and retrospective topic classification over static LDA and PageRank-only baselines, while providing interpretable parameters that reveal persistent excitation from social-stream and biomedical enrichment toward ontology/linked data. We release reproducible code and provide two figures (topic shares and excitation matrix) and two summary tables (performance and top excitation links) to support replication and extension.