GSP (Geo-Semantic-Parsing): Geoparsing and Geotagging with Machine Learning on Top of Linked Data

摘要: Recently, user-generated content in social media opened up new alluring possibilities for understanding the geospatial aspects of many real-world phenomena. Yet, vast majority such lacks explicit, structured geographic information. Here, we describe design and implementation a novel approach associating information to text documents. GSP exploits powerful machine learning algorithms on top rich, interconnected Linked Data order overcome limitations previous state-of-the-art approaches. In detail, our technique performs semantic annotation identify relevant tokens input document, traverses sub-graph extracting possible related identified optimizes its results by means Support Vector Machine classifier. We compare with those 4 techniques baselines ground-truth data from 2 evaluation datasets. Our achieves excellent performances, best \(F1 = 0.91\), sensibly outperforming benchmark that achieve \le 0.78\).