A list of recent papers about Graph Neural Network methods applied in NLP areas.

A modern multimodal knowledge graph with type-specific metadata across biomedical domains.

This project focuses on predicting maternal-fetal health outcomes, specifically early detection of preeclampsia, by learning transferable representations from cfRNA and placental transcriptomic data.

Infra2Graph prototype for bridge impact assessment in city scale. The system integrates heterogeneous graph analysis and HGNN-based closure-impact prediction, with Exp-4-lite (kNN, no edge attributes, weak quantile-based weighting) adopted as the reference MVP configuration.

Time-Relaxed Directed GNN for Bitcoin Fraud Detection | 6 Novel Contributions | Production-Ready | E7-A3: 0.5846 PR-AUC (+4.1%) | E9 Fusion: +33.5% | Publication-Ready Research

ConHGNN-SUM: Contextualized Heterogeneous Graph Neural Networks for extractive document summarization. Published in IEEE AISP 2024. Revolutionizes summarization by modeling documents as dynamic graphs with semantic relationships between words and sentences.

Implementation of a Heterogeneous Temporal GNN (HTGNN) for DeFi anomaly detection. Includes extensive baseline comparisons (RGCN, GAT, XGBoost), ablation studies, and a Neo4j-backed inference pipeline.

PyTorch implementation of Heterogeneous Graph Neural Networks with attention aggregation for node classification on academic networks.

A case study repository demonstrating the applicability of Bridge Importance Scoring v1.6 to other municipalities beyond Yamaguchi City. This repository validates the system's transferability by applying the complete v1.6 pipeline (heterogeneous graph analysis + HGNN-based closure-impact prediction) to Tama City, Tokyo.

🕐 Uncover Bitcoin fraud with our zero-leakage temporal GNN, designed for effective detection through a systematic, experimental approach.