Hi Erik, Interesting project, though I must admit some caution about its success. How do you plan to develop readership for this site? Yelp seems to have a commanding lead.
Pine On Aug 5, 2016 18:17, "Erik Moeller" <eloque...@gmail.com> wrote: > Hi Michał, > > Thanks for your comments! > > On Fri, Aug 5, 2016 at 4:55 PM, Michał Brzozowski <www.ha...@gmail.com> > wrote: > > > Have you devised any robust algorithm for linking OSM primitives to > > objects in the external database? In general case, it seems really > > hard to track objects as they get converted from nodes to areas, or > > decide whether given OSM feature is no longer representing some entity > > in the external database. > > No, and I'm not very familiar with OSM's data structures and APIs yet. > What I'm imagining for now as the initial OSM-related features are: > > 1) enabling search for POIs similar to http://openpoimap.org/ but more > lightweight and purpose-focused (so you can start a review and just > select a POI from the map to identify it) > > 2) importing (and attributing!) relevant data on demand, which by the > looks of e.g. https://www.openstreetmap.org/way/422293736 seems like > it often includes quite a bit of relevant data that future reviewers > would appreciate. > > If possible, I'd also like to add: > > 3) flagging imported data as read-only and synchronizing it in regular > intervals. People who want to improve that data would then just be > pointed to OSM (or Wikidata, or whatever community source). > > I have no intention of performing a bulk import anytime soon; while > this could be good for bootstrapping, it will be too big of a > technical challenge too early, I think. Instead for now we'll > add/import metadata about things we review if/when we review them. > > Do you see fundamental technical challenges with any of the above? I > don't think conversion from nodes to areas would necessarily be > problematic, as long as the sync job can learn that such a change has > occurred to the object it's trying to keep in sync. > > > A framework / API for performing such linking would be of great > > interest, as it could enable many applications to exist on top of OSM > > - recognizing that not everything belongs to OSM. > > *nod* OSM-land is interesting compared with the Wikimedia world I'm > more familiar with, with much more emphasis on a large distributed > community building tools and APIs, some proprietary, some open. I'll > want to look at the state of the open tools out there to see if what > I'm describing above can already be built, or if there's someone who's > willing to collaborate! > > > Regarding the idea, I reckon it may not scale well, if at all. Weeding > > out spammers needs constant attention, and community moderation is > > prone to the Sybil attacks. This may be less of a problem on sites > > such as OSM or Wikipedia where data needs verifiability that or > > another way (so in order to gain trust you have to do actual work). > > Reviews are inherently subjective. Not to mention any legal BS one may > > get from business owners. > > Heh, it's certainly a hard problem. :) Here are a few things to note: > > - Currently the system is invite-only and likely will be for a while. > I reckon building a core community that cares about quality, > organization, etc. will take a while, and we can then give a lot of > those folks permission to also act as moderators so they can ban > spammers once we (temporarily or permanently) open the floodgates. > Invitation is something we can give away liberally, but it functions > as a bit of a barrier to entry for bad faith actors. > > - I'm building into the architecture strong notions of trust and > affiliation. Users can be members of like-minded teams with given > rules (think sub-reddit as an analogy), and they can individually > express trust toward one another, so we can track the trust graph that > allowed an abuser to act with elevated trust levels. Trust will likely > factor into ranking calculations, visibility of content, and so on. To > give an example, it's already the case that the reviews shown on > https://lib.reviews/ are written by users with the "trusted" flag set, > while https://lib.reviews/feed shows all (unfiltered) reviews. > > - In general, my experience with Wikimedia has taught me that > transparent community collaboration in good faith is a pretty good way > to deal with such problems. Wikimedia has to deal with paid PR flacks > regularly, for example, and generally has established procedures for > spotting and kicking out such folks. Similarly, WMF has had to face > down nasty legal threats long before it had a big budget. As long as I > give the community good tools to self-organize rather than following > an enterprise-style approach of solving everything from the top down, > I am optimistic that we can make decisions such as "when do we open > the floodgates" collaboratively.) > > Warmly, > Erik > > _______________________________________________ > talk mailing list > talk@openstreetmap.org > https://lists.openstreetmap.org/listinfo/talk >
_______________________________________________ talk mailing list talk@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk
