Multi-Hop Question Generation with Knowledge Graph-Enhanced Language Model

The task of multi-hop question generation (QG) seeks to generate questions that require a complex reasoning process spans multiple sentences and answers. Beyond the conventional challenges what ask how ask, QG necessitates sophisticated from dispersed evidence across sentences. To address these challenges, knowledge graph-enhanced language model (KGEL) has been developed imitate human for questions.The initial step in KGEL involves encoding input sentence with pre-trained GPT-2 obtain comprehensive semantic context representation. Next, graph is constructed using entities identified within context. critical information related answer then utilized update representations through an answer-aware attention network (GAT). Finally, multi-head module (MHAG) performed over updated latent coherent questions. Human evaluations demonstrate generates more logical fluent compared GPT-2. Furthermore, outperforms five prominent baselines automatic evaluations, BLEU-4 score 27% higher than