Question Answering on Linked Data: Challenges and Future Directions Saeedeh Shekarpour Knoesis Research Center, USA saeedeh@knoesis.org Dennis Lukovnikov University of Bonn, Germany lukovnikov@outlook.com Ashwini Jaya Kumar Fraunhofer IAIS, Germany ashwinijk18@gmail.com Kemele M. Endris University of Bonn, Germany endris@cs.uni-bonn.de Kuldeep Singh Fraunhofer IAIS, Germany kskuldeepvit@gmail.com Harsh Thakkar University of Bonn, Germany thakkar@cs.uni-bonn.de Christoph Lange University of Bonn / Fraunhofer IAIS, Germany langec@cs.uni-bonn.de ABSTRACT Question Answering (QA) systems are becoming the in- spiring model for the future of search engines. While re- cently, underlying datasets for QA systems have been pro- moted from unstructured datasets to structured datasets with highly semantic-enriched metadata, but still question answering systems involve serious challenges which cause to be far beyond desired expectations. In this paper, we raise the challenges for building a Question Answering (QA) sys- tem especially with the focus of employing structured data (i.e. knowledge graph). This paper provide an exhaustive insight of the known challenges, so far. Thus, it helps re- searchers to easily spot open rooms for the future research agenda. Keywords Question Answering System, Research Challenge, Speech Interface, Query Understanding, Data Quality, Distributed and Heterogenous datasets, Interoperability of Components. 1. INTRODUCTION Web of Data is enormously growing (currently more than 84 billion triples 1 ). This data contains both structured and unstructured data. Still, taking advantage of this rapidly growing data is challenging. Traditional information re- trieval approaches based on keyword search are user-friendly but can not exploit the internal structure of data due to their bag-of-words semantic. For searching information on 1 observed on 14 October 2015 at http://stats.lod2.eu/ ACM ISBN 978-1-4503-2138-9. DOI: 10.1145/1235 the Data Web we need similar user friendly approaches i.e. keyword-based interfaces, which lie on the internal structure of the data. Question Answering is a specialized form of information retrieval. A Question Answering system retrieves exact an- swers to questions posed in natural language by user. While recently, underlying datasets for QA systems have been pro- moted from unstructured datasets to structured datasets with highly semantic-enriched metadata, but still question answering systems involve serious challenges which cause to be far beyond desired expectations. Question Answering systems consists of elements which in- dependently can be studied and developed. These elements consists of (1) input interface for obtaining query, (2) under- standing, interpreting, disambiguating and parsing query, (3) issues related to the employed datasets such as hetero- geneity, quality and indexing and (4) interoperability issue for interacting different components. In the following, we elaborately discuss challenges related to each element and possibly future directions which can be considered. We close with the conclusion and future plan. 2. CHALLENGES In this section we present question answering challenges from four different aspects namely, (i) Speech-based inter- face challenge, (ii) query understanding, interpreting, dis- ambiguating and parsing challenges, (iii) data-oriented chal- lenges (iv) interoperability of QA components challenge. 2.1 Speech Interface Interfacing speech to QA systems has become a focus of research for a long time. But, the main focus of research effort so far has been spent on interfacing speech to IR-based QA systems.[41, 43, ?], and much less on interfacing speech input to KG based QA systems. Typical state-of-the-art IR approaches integrate a speech recognition (SR) unit directly with the QA system. An effort beyond merely interfacing the two units is required to enable natural conversation in question answering system for both IR and KG methods. An SR system mainly consists of an acoustic model and a