Options
2019
Conference Paper
Title
Offline question answering over linked data using limited resources
Abstract
Question Answering over Linked Data provides concise information to the user from a natural language request instead of ooding them with documents. However, the accessibility of Linked Data resources, e.g., SPARQL endpoints, is bound to an online connection. We present OQA, the first offline Question Answering system over Linked Data for mobile devices. We built OQA with the limited resources of an Android mobile device, such as battery power, computational power, or memory consumption in mind. Our OQA system has three main components: 1) question analysis and 2) query generation which identify the type of the question and reform it into a semantically meaningful data structure, i.e., a SPARQL query. Finally, the 3) query execution uses a novel mobile triple store, implemented with RDF4J. Our evaluation suggests that OQA is feasible for daily use in terms of battery consumption and able to answer domain-specific questions with up to 72% accuracy.