[scary-size]

Interesting, does the server fetch the html itself? Or does the client POST it? The letter would be useful for paywalled sites like Cook‘s Illustrated. I spent quite some time on ingredient labelling (what’s a unit, quantity etc.). Feel free to ping at hn[at]franz.hamburg

[thebricklayr]

> does the server fetch the html itself?

Yep, the client just sends the recipe URL and then the server fetches the HTML. Agreed, it would be better to have the client send the HTML for paywalled sites, which is another reason I just want to do it all on the client.

> I spent quite some time on ingredient labelling (what’s a unit, quantity etc.)

I can relate to you there. It was a long process of trial and error for me to get right, and there are still plenty of edge cases left to handle. Long-term I think AI + NLP will make this kind of thing easier, but for me it wasn't fast, reliable, cheap, or portable enough to run in an iOS app in real time quite yet.

[scary-size]

Nice! I dabbled with deep learning and CRF for the sequence tagging. I ran in the same issues you mentioned. Current approach is hand rolled. The „holy grail“ would be linking ingredients mentioned in the directions. So, I could just tap on „…vinegar…“ in an instruction and see that I need 2 tbsp.