Thanks! This is exactly why I'm looking for a feedback. These are real errors from an automated OSM import that got hallucinated descriptions. We've corrected them:
- Joella's Hot Chicken → now removed "Hot chicken" (unless it's vegan replacement) - has no place on the platform, especially if it's a chain. Even if it has a vegan option (I'm preparing a "last chance" filter so that person may see the chains, maybe if there is no other option around).
- Jasmine Thai → now "Vegan-Friendly" (not 100% vegan)
- 10th Street Diner → upgraded to "100% Vegan" where it belongs
You're right that the AI classifier made embarrassing mistakes on these. We have a community correction system (tap any place -> "Suggest Correction" -> you can change the vegan level), and we're running a bulk re-classification pass to catch more of these. Also I've added the "Actually 100% vegan" button to the list of available verifications.
Overall I'm improving the scraping and enrichment scripts on daily bases thanks to reports like this. I'm quite sceptical about current data state, so reports like this are extremely helpful! Appreciate it!
Contributing back is something I'd genuinely like to do - the challenge is that bulk OSM imports require a formal community proposal, import account, and changeset review process, so it's not as simple as a one-way sync - will work on that once will be more confident on the data quality. That said, individual fixes you add to OSM will flow into PlantsPack on the next sync.
For the fields that matter most:
If you're updating OSM entries for vegan restaurants, diet:vegan=only is the highest-impact change you can make - that one tag is enough for PlantsPack to promote a place to the fully-vegan tier on next sync. We have 10+ steps verifications and enrichment process for each place, so even if places is not fully described on OMS - it may end up being better represented on our platform + we run extra checks to verify its vegan_level and status with Google API and AI checks.