There was a really nice post on doing this kind of thing with CRF back in 2015 [0]. Open source data, and code on github. Also a nice tutorial on structured pre

There was a really nice post on doing this kind of thing with CRF back in 2015 [0]. Open source data, and code on github. Also a nice tutorial on structured prediction using CRF type models. Would be interesting if you could prompt, LoRA distill, or use modern LLM tricks against a well-labeled and curated set, similar to how other tagging problems are handled with modern pretrained models. [0] https://open.nytimes.com/our-tagged-ingredients-data-is-now-on-github-f96e42abaa1c https://open.nytimes.com/our-tagged-ingredients-data-is-now-...