[dunham]

Years ago, I wrote something like you describe in the blog (regex to match ingredient lines, looking for imperative verbs, filling in the gaps). Recently I revisited the subject and learned that almost everybody has decent jsonld data now. Even paywalled stuff.

Now I've got tampermonkey watching over my shoulder and backing up everything I look at to a couchdb instance. (Still gotta write some UI and an agent to pull down images, but I've got other irons in the fire at the moment.)