Hierarchical Graph Convolutional Network Approach for Detecting Low-Quality Documents

A hierarchical graph convolutional network (HGCN) identifies document quality by detecting internal semantic inconsistencies and title-to-body incongruences. The model utilizes a new Inconsistency Dataset derived from news metadata to map relationships between textual components. This architecture enables automated detection of low-quality content that lacks structural or topical coherence.