Enabling downstream machine-learning over the textual information contained in building knowledge graphs

Abstract

The current practice of statistical learning from building information models mostly relies on the manual construction of table-oriented representation of the numeric data. This is while the lexical information contained in building information models can further reinforce the learning process by preserving the essential role of the semantic relationships. In this paper, application of one of the state-of-the-art knowledge graph embedding algorithms (RDF2Vec), yielded promising results with regards to an object clustering task. This paper contributes to the literature by shedding light on the potential of semantic embedding algorithms to facilitate downstream machine learning over building knowledge graphs.