Max Bachmann

The results of 100 is correct and the same result is returned by fuzzywuzzy as well. I agree this could probably be documented better. Improvements to the doc are welcome....

I reformatted your post to make it a bit more readable. > According to your detailed explanation, Currently fuzz.token_set_ratio has no difference with fuzz.partial_token_set_ratio. It does behave in a different...

> Please don't break the api. I am not quite sure what your are talking about, This would not break the API in any way. Unless you rely on the...

This appears to be OSA and not Damerau-Levenshtein.

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7724867/ sounds cool but requires a small alphabet. The `last_row_id` array could be a fast hashmap, which is worth testing. Either this or Hirschberg should be implemented for long sequences...

I took a first stab at implementing the Damerau Levenshtein distance without any weights (all operations have weight 1). Here are the current benchmark results:   In addition this...

simd leads to a large improvement in performance: Old ``` BM_LCS 93868375 ns 93704140 ns 7 InvRate=36.6032ns Rate=27.32M/s BM_LCS 122188164 ns 121960772 ns 6 InvRate=47.6409ns Rate=20.9904M/s BM_LCS 176089869 ns 175789716...

CMake support got added now

@jamesturk is this still an issue? as far as I can see this is fixed: ```python >>> jellyfish._jellyfish.match_rating_codex('ä') 'Ä' >>> jellyfish.cjellyfish.match_rating_codex('ä') 'Ä' ```

https://github.com/rhasspy/rhasspy-server-hermes/pull/15 is a related PR. Right now the wake button sets a session id, that is invalid. I remember, that I asked @synesthesiam about this behaviour back than aswell. However...