Abstract: Meta-paths are widely adopted in heterogeneous information networks to delineate relationships between different entities. While existing heterogeneous graph neural network models predominantly utilize meta-paths to uncover complex relationships within the network, some studies have ventured into using solely edge type information for representing inter-entity relationships. However, these methods have limitations as they over-rely on meta-paths or place undue emphasis on edge types, failing to fully harness the synergistic potential of meta-paths and edge types. A novel heterogeneous graph neural network model HGME is proposed, which integrates both meta-paths and edge type information for the first time, synergistically processing relationships among entities in heterogeneous information networks through meta-path and edge type enhancement approaches. The model's efficacy is assessed using three heterogeneous information network datasets: ACM, DBLP and Freebase. The results demonstrate that the HGME model outperforms traditional heterogeneous graph neural network models in classification and clustering tasks across three datasets. The effectiveness of HGME is further corroborated by visual analysis.