Neli Zlatareva, Devansh Amin

Abstract: SPARQL is a powerful query language for an ever-growing number of Semantic Web applications. Using it, however, requires familiarity with the language which is not to be expected from the general web user. This drawback has led to the development of Question-Answering (QA) systems that enable users to express their information needs in natural language. This paper presents a novel dependency-based framework for translating natural language queries into SPARQL queries, which is built on the idea of syntactic parsing. The translation process involves the following steps: extraction of the entities, extraction of the predicate, categorization of the query’s type, resolution of lexical and semantic gaps between user query and domain ontology vocabulary, and finally construction of the SPARQL query. The proposed framework was tested on our closed-domain student advisory application intended to provide students with advice and recommendations about curriculum and scheduling matters. The advantage of our approach is that it requires neither any laborious feature engineering, nor complex model mapping of a query expressed in natural language to a SPARQL query template, and thus it can be easily adapted to a variety of applications.

Keywords: Information Retrieval, Natural Language Processing, Semantic Web, SPARQL, Question-Answering Systems.

Date Published: November 10, 2021
DOI: 10.11159/jmids.2021.006

View Article