The question in general is "Yes, RediSearch supports all of these features". You can read about it all in the docs https://redisearch.io
I also pointed bellow to the specific relevant area in the docs.
> 1. What human languages does it support. > 2. In these human languages how does stemming and decompounding work in your implementation.
https://oss.redislabs.com/redisearch/Stemming/
> 3. how is word importance determined in your index - TF-IDF? other algorithm? Are least important words automatically dropped from queries? >4. Do you have ability to rank on both the stemmed/decompounded query/results and exact matches? So something like raw field access.