Treo: Combining Entity-Search, Spreading Activation and Semantic Relatedness for Querying Linked Data

This paper describes Treo, a natural language query mecha- nismfor Linked Data which focuses on the provision of a precise and scalable semantic matching approach between natural language queries and distributed heterogeneous Linked Datasets. Treos semantic matching approach combines three key elements: entity search, a Wikipedia-based semantic relatedness measure and spreading activation search. While entity search allows Treo to cope with queries over high volume and distributed data, the combination of entity search and spreading activation search using a Wikipedia-based semantic relatedness measure provides a flexible approach for handling the semantic match between natural language queries and Linked Data. Experimental results using the DBPedia QALD training query set showed that this combination represents a promising line of investigation, achieving a mean reciprocal rank of 0.489, precision of 0.395 and recall of 0.451.