Persian Question Answering over Linked Data

Tracking #: 2043-3256

This paper is currently under review
Mitra Isaee
ahmad zaeri1
Mehdi Jabalameli
Mohammadali Nematbakhsh
afsaneh fatemi

Responsible editor: 
Philipp Cimiano

Submission type: 
Full Paper
The purpose of linked data is to publish data on the web so that they can be interpreted and read by machines. Structured query languages such as SPARQL provide access to structured data in linked data, but web users are not interested in writing formal query languages given their difficult syntaxes. One solution to this problem is the use of natural language interfaces. In recent years, researchers have focused on the issue of developing natural language interfaces for linked data; however, research on developing a Persian question answering system for linked data is limited. In this study, a system is proposed to convert a Persian natural language query to SPARQL instructions. The system uses existing tools for the linguistic analysis of Persian questions, as well as mapping natural language expressions to DBpedia ontology concepts. Furthermore, novel algorithms are used to create a disambiguation graph from the results of the analysis step and to transform a Persian question into a SPARQL query using the disambiguation graph. Next, the system executes the constructed SPARQL query against DBpedia and extracts the results. The system’s performance is evaluated and compared to that of other systems, with the results revealing the superiority of the proposed system.
Full PDF Version: 
Under Review