The continuous need to process semi-structured data in the more connected and semantic web requires a retrieval model that can truly reflect the user's intention and capture a user's understanding. As a semantic network shows great potential in representing the inherent structure of information in a document, recent studies have attempted to apply semantic networks into information retrieval. While many of the recent works on semi-structured data retrieval focused on the use of field structure within the data. Solely relying on the field structure is insufficient to portray the user's understanding, which is represented through the use of specific query terms. In this study, we seek to overcome this limitation by utilizing a semantic network to model semi-structured data and apply a graph-based semi-structured data retrieval model. Using both a popular testing environment and a real-life query data, we compare the performance of the suggested model with various competitive state-of-the-art retrieval models. The study's findings demonstrate the strength of the proposed model while providing intriguing opportunities for further application of the model.