Information Extraction from a Large Knowledge Graph in the Nepali Language

Abstract

Information is abundant in the web. The knowledge graph is used for organizing information in a structured format that can be retrieved using specialized queries. There are many Knowledge graphs but they differ in their ontologies and taxonomies as well as property types that bind the relation between the entities, which creates problems while extracting the knowledge from them. There is an issue in multilingual support. While most of them claim to be multilingual they are more suitable for querying in the English language. Most of the existing knowledge graphs in existence are based on Wikipedia Info box. In this work, we have devised an information extraction pipeline for retrieving knowledge in Nepali Language from Wikidata using SPARQL endpoint. Queries based on Wikipedia info box has more accurate responses than the Queries based on the paragraph content of Wikipedia articles. The main reason behind that is that the information inside the paragraph is not linked properly in the Wikipedia info box.