Reducing the impact of out of vocabulary words in the translation of natural language questions into SPARQL queries

Accessing the large volumes of information available in public knowledge\nbases might be complicated for those users unfamiliar with the SPARQL query\nlanguage. Automatic translation of questions posed in natural language in\nSPARQL has the potential of overcoming this problem. Existing systems based on\nneural-machine translation are very effective but easily fail in recognizing\nwords that are Out Of the Vocabulary (OOV) of the training set. This is a\nserious issue while querying large ontologies. In this paper, we combine Named\nEntity Linking, Named Entity Recognition, and Neural Machine Translation to\nperform automatic translation of natural language questions into SPARQL\nqueries. We demonstrate empirically that our approach is more effective and\nresilient to OOV words than existing approaches by running the experiments on\nMonument, QALD-9, and LC-QuAD v1, which are well-known datasets for Question\nAnswering over DBpedia.\n