Unsupervised representation learning can train deep learning models to formally express the semantic connotations of objects in the case of unlabeled data, which can effectively realize the expression of the semantics of geospatial entity categories in application scenarios lacking expert knowledge and help achieve the deep fusion of geospatial data. In this paper, a method for the semantic representation of the geospatial entity categories (denoted as feature embedding) is presented, taking advantage of the characteristic that regions with similar distributions of geospatial entity categories also have a certain level of similarity. To construct the entity category embedding, a spatial proximity graph of entities and an adjacency matrix of entity categories are created using a large number of geospatial entities obtained from OSM (OpenStreetMap). The cosine similarity algorithm is then employed to measure the similarity between these embeddings. Comparison experiments are then conducted by comparing the similarity results from the standard model. The results show that the results of this model are basically consistent with the standard model (Pearson correlation coefficient = 0.7487), which verifies the effectiveness of the feature embedding extracted by this model. Based on this, this paper applies the feature embedding to the regional similarity task, which verifies the feasibility of using the model in the downstream task. It provides a new idea for realizing the formal expression of the unsupervised entity category semantics.

Geospatial data contains abundant temporal, spatial, and attribute information related to geographic entities, serving as a critical foundation for the development of new quality productive forces. Achieving effective semantic representation of geographic features is a key prerequisite for enabling intelligent data analysis and application. Spatial relationships are an essential semantic characteristic of geographic features and are among the most commonly introduced characteristics for representation learning in recent studies. However, current research primarily focuses on point-based features (such as POIs) or on converting line and polygon features into point features, leading to inaccuracies in representing spatial proximity between features, which in turn affects the representation outcomes. In this paper, a feature unit delineation method based on geometric shapes is proposed to better construct the proximity relationship graph between features. It enriches the semantic features of feature units using a large language model and trains an edge optimization model to refine the graph structure. Building upon iGCL, a semantic representation learning model for geographic features is constructed. Finally, using Xiamen Island and Nanchang City as study areas, experimental datasets are built by integrating OSM and Amap (Gaode) POI data. The performance of the proposed model is validated and compared with other models through three tasks: semantic overlap calculation, node classification, and node clustering. The results show that the feature representations obtained by the proposed model achieve the best performance across all tasks, fully demonstrating its effectiveness and superiority in geographic feature representation learning, thereby providing a foundation for subsequent intelligent geographic feature tasks.