@reiver@reiver ⊼ (Charles)
very nice!
#FediDev
considering "from scratch" to be the default ;) I have written today the generator for semantic compression in any language.
What I am working on is a compressor for ActivityPub Objects with Natural Language Maps. It converts anything to a UInt8Array which is only 20-25% of the original.
Basically it loads a news corpus from the Leipzig Uni (not the famous Leipzig HGB Academy Of Visual Arts, from the ordinary University but it's fine ;) Then it generates a definition for the most 256 common languages.
Made a demo for the before (from scratch generated) DE 2 days ago https://github.com/sebilasse/compressDE and it will be combined with the language detection of 852 locales https://github.com/redaktor/languages