How to Tune Parameters in Geographical Ontologies Embedding

Many Natural Language Processing (NLP) tasks, like question answering or analyzing verbatim comments, have started to use word embeddings due to their ability to capture semantic relations between words. Recently, embeddings have been also applied in the geospatial context to represent geospatial ontologies, thanks to their ability to capture semantic similarity. In this paper, we present an analysis of a promising embedding technique particularly suitable for representing hierarchical structures. We conduct a deep technical evaluation of many parameters and their impact on the quality of the representation.