Create README.md

I provide data scrapped from askimam.org
The json files contain 24081 questions, metadata and embeddings of answers.
I provided embeddings for the english and the arabic part of the answer in addition to embeddings of every sentence in the answer. I use truncation when the length of the answer is too large (embeddings of sentences are usually within the length limit).
The embeddings were computed using the following models
- bert-base-uncased (english parts)
- asafaya/bert-base-arabic (arabic parts)
- bert-base-multilingual-uncased (english and arabic parts)

README.md

@@ -0,0 +1,8 @@
+---
+language:
+- en
+- ar
+pretty_name: askimam
+size_categories:
+- 10K<n<100K
+---