Key points are not available for this paper at this time.
The knowledge bases required the query language SPARQL, which consists of subject, property, and object. SPARQL is a structured query language and is difficult to understand. That issue becomes a problem in natural language processing queries. One situation in question answering is how to translate natural language into a structural SPARQL. This work aims to develop an automated SPARQL template algorithm regardless of the pattern structure of the query triples. It provides a more varied SPARQL query for data retrieval named Flexible SPARQL. This approach initially lies in combining elements of RDF with basic techniques of natural language processing to generate a template of SPARQL. In this work, the approach to making automatic templates is proposed without regard to the pattern of the triple structure or the location of the subject and object. Template-based research that exists today still uses rules to determine the position of subjects, objects, and properties in the SPARQL structure. Therefore, this work used the QALD 7 question set and DBpedia dataset. The previous systems utilized the same questions and data sets. Despite the simple proposed approaches that do not use complex, sophisticated techniques, they have shown promising results compared to the previous systems. The accuracy result from 215 questions is 73% and micro-Recall 0.701, micro-Precision 0.664, micro-F-Measure 0.682, macro-Recall 0.711, macro-Precision 0.592, macro-F-Measure 0.646. Overall, the Flexible SPARQL system has higher results on several measurements that define a promising approach. However, it's important to note that Flexible SPARQL generally tends to fail at generating complex SPARQL, which is a limitation of the system.
Wardani et al. (Mon,) studied this question.