Upload askimam_.tar.xz

Contains 97 json files of around 250 questions each. The dataset contains 24081 questions from askimam.org
It contains metadata and answer embeddings. The answers in askimam have an english and an arabic part. We generated
Bert embeddings using bert-base-uncased and bert-base-multilingual-uncased for the English part and bert-base-multilingual-uncased and asafaya/bert-base-arabic for the Arabic part.
This required truncating the answers to fit the Bert context.
We also generated an embedding per sentence using the same models. These sentences were obtained using the html formatting of askimam that organizes answers into sentences.

askimam_.tar.xz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:29ee8adf37ab34a6e862db9274d7d0b6d579e727c350266121df438ac2bd23a4
+size 304525148