Techniques, datasets, evaluation metrics and future directions of a question answering system

Question answering has been around for more than half a century. The problem was addressed with different solutions in the eras of different technologies. Some proved more helpful and accurate than the other. Different studies are available online which list and summarize the work done in this domain. This SLR adds up to that list with answers to some questions which will assist the researchers in this field to comprehend the existing knowledge, quickly analyze the available facts and determine some research gaps and future directions. In this article, we investigate different solution domains applied to question answering systems, their results, and methodologies. We also list and discuss different datasets provided to the community for experiments along with their availability status. In the light of this study, we analyze different solution domains and the areas where they produce promising results. Moreover, we focused on different evaluation metrices used in the papers that were included in this study and shed light on some metrices which should be included in the results if the community wants to achieve greater results. Lastly, we also looked into an interesting possibility of a question answering system where answer could be generated using multiple sources. And for that we suggested a domain based on the Quran, Tafseer and Ahadith data sources as the Quran and Ahadith contribute collectively in the Islamic legislation. We hope this article will help the new researchers in the field of question answering to start their research.