(disclaimer: I work on Duckling)

I don't know of any benchmarks. HeidelTime and SUTime are two solid projects... although the rule files [0][1] are a bit scary if you ask me :-).

Quick thoughts:

- SUTime relies on TokensRegex [2] which is similar to how Duckling parses sentences at a high-level

- SUTime seems to only provide English rules

- I don't know of any production use cases of either

[0] https://github.com/stanfordnlp/CoreNLP/tree/master/src/edu/s...

[1] https://github.com/HeidelTime/heideltime/blob/master/resourc...

[2] https://nlp.stanford.edu/pubs/tokensregex-tr-2014.pdf