Leveraging multilingual descriptions for link prediction: Initial experiments

In most Knowledge Graphs (KGs), textual descriptions ofentities are provided in multiple natural languages. Additional informa-tion that is not explicitly represented in the structured part of the KGmight be available in these textual descriptions. Link prediction modelswhich make use of entity descriptions usually consider only one language.However, descriptions given in multiple languages may provide comple-mentary information which should be taken into consideration for thetasks such as link prediction. In this poster paper, the benefits of mul-tilingual embeddings for incorporating multilingual entity descriptionsinto the task of link prediction in KGs are investigated