WikiQAar

WIKIQAar is a bilingual English--Arabic Question Answering corpus built on top of WIKIQA. In order to build WIKIQAar, we independently produced two (sub-)corpora. On the question side, we produced a parallel corpus by translating the questions into Arabic. We applied two automatic machine translators and crowdsourced the selection of the best one to be incorporated into the corpus. On the reference side, we produced a comparable corpus by retrieving the Arabic edition of the corresponding Wikipedia articles. In order to identify the exact answers, we applied a supervised model to find out which text fragment in the Arabic article corresponded to the answer in English (hence composing yet another parallel collection).
We make this dataset publicly available for the community to explore models such as cross-language question answering and answer triggering across languages; i.e. deciding if an answer exists in the same language as the question and ---if not--- searching in the other language.