An attempt to compare results of

• https://github.com/blevesearch/segment
• https://github.com/clipperhouse/uax29

Have a look at the test file.

What I've found (June 2022)

• Bleve splits a run of spaces into separate tokens, while uax29 returns a single token with multiple spaces
• Bleve appears to be Unicode 8.0.0, uax29 is 13.0.0, seems like there's a difference on emoji skin tone modifiers, which might generalize to emoji modifiers in general?

Both the Bleve segmenter and UAX29 pass the Unicode test suite (for their respective Unicode versions).