- is based on OSM
- has a review feature like Google Maps
- is not tied to a company
- is open source (optional)
That seems like properly hard problem - reviews aren't objective facts like OSM data & verification/spam would be tricky.
Now, however, I worry that Google will try to monetize the maps more and more in ways that are hostile to the user. They have already started showing ad-sponsored restaurants/locations on the maps, last I checked. Reviews are so critical to many businesses (particularly restaurants), it seems important for maps and reviews to be run by a user-friendly entity, ideally something like OSM.
Providing such a service is expensive, however. The fact that OSM does not have a decent Android app even is telling (Osmand is laggy and barely usable; Organic maps is nice, but is not a replacement for Google maps).
So we are stuck in the situation we are in, and I am not sure what a good solution would be.
* Allow anyone to review.
* When viewing other people's reviews, you rank them by how similarly they review places you and they have both been to. Ie. Do you agree with them? If you agree with them, then weight that users reviews higher. If they rank places highly that you think got 1 star, then rank them negatively.
* This approach can be done on the complete review graph, so you don't need to have reviewed the same places someone else ranked.
* The downside is you can't see any meaningful reviews till you write a few yourself.
Obviously any such scheme is helped by there being not-too-many bot accounts. It's hard to limit bot account creation in a decentralized opensource design.
One potentially workable approach is to have users leave a trail of data in some immutable form (eg. a series of reviews left, with scores, timestamps, and text comments), which you later apply some heuristic over (eg. 'all comments containing a web link to a url shortener are probably spam'). All data that is older than the publication of the heuristic is then fairly trustworthy, since bot account owners can't retrospectively edit their spam to make it not match the heuristic.
Have each reviewable location publish a key on a sign in their establishment. Make reviewers scan a QR code and sign their comment with that key.
That ensures that while the key doesn't leak every review must be written by someone who has physically visited.
Obviously you need to trust every physical visitor to not publish the sign, but I suspect that might be sufficient to stop spam nonetheless.
A recommended alternative is Mangrove [1] which allows anonymous cryptographically verified reviews, but I have neither had a closer look yet nor given it a try so far.