[meeting] [joint] 31.05.2024

[deemp]

@fizruk's notes

Meeting 2024-05-31

1. Closing Stage 1

• We have fast version without elaboration, works well on most tests.
• We have instructions for the pipeline, is it enough to reproduce everything? @yegor256
• Do we need the metrics to be perfect for Stage 1? @yegor256
  • If yes, there is a problem:
  • Formations metric is technically impossible to improve on the given tests. We should either have different tests, or compute metrics differently. https://github.com/objectionary/normalizer/issues/380
  • Applications and dispatch is reduced, but not as much as needed on some tests (>85% tests are good, but we need ≥90%)

2. ITP 2024 paper is rejected:

• Reviews are positive, but identify incomplete work.
• Need to select another target for a revised version of the paper.

3. Spring example pipeline

• Do you have a preliminary sketch of the idea for the transformation? @maxonfjvipon
• We plan to experiment with the current collection of phi files

4. Performance of the normalizer:

• It is slow, and there are many opportunities to optimize its performance:
  • Using withSubObjects causes severe slowdown, need to investigate
  • We should implement a version of NbE for truly efficient normalization
  • One significant slowdown that I'm not sure how to improve properly: choosing the minimum term in a chain

5. Normalizer paper:

• After closing Stage 1
• I have an idea for a (generalized) two-phase rewrite-based framework, but we need to test it first
• I also would like to know if we need a separate paper for Stage 2, or ITP paper covers it

6. Normalizer patent:

• The idea is to specify architecture and concrete rewrite-based rules on top of the method outlined in the Normalizer paper
• After closing Stage 1