-
Notifications
You must be signed in to change notification settings - Fork 74
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Add option to diff conflate to not treat reviews as matches #3387
Comments
@curranMapper This is happening because there is a single restaurant in the NOME layer which is triggering reviews against every POI in that area of the OSM dataset. Since all matches get dropped by diff conflate, we lose all of those OSM restaurants. POI conflation is pretty lenient on reviews in general and the match distance fairly large. So, Monday I'll look into making the POI matching more strict for this use case, as the leniency is causing features to get dropped from the output. |
In thinking about other ways user could work through this, advanced options has some "distance thresholding" settings that can be manipulated. This seems to be for POI to polygon situations only, and not POI to POI. |
From what I mentioned, 1) is time consuming to implement, 3) is very time consuming to implement, but 2) wouldn't take too long. Not sure which of the options (or maybe all of them?) are the best idea at this point. |
@curranMapper What do you think? |
Hey @bwitham, my input below.
I do see utility to no 2 as option, however, would this behavior be unique to POIs only? It would pass along POIs normally flagged as review but wouldn't pass through roads flagged for review as example? Also considering benefit of keeping behavior consistent across different features so I would vote for seeing more use cases highlighting this as "issue" to justify benefit of adding no 2 to the mix and keep with 3 as long term goal in order to keep behavior consistent across features. Will also bring this up to users as applicable to see if need develops to have no 2 implemented. Does that make sense to you? others thoughts on this? |
For 2), it could be done either way. Either just pass through POIs marked for review or pass through everything marked for review. Yeah, there's no point in adding 2) unless we're sure it provides utility, as it will just add to complexity in the app if it doesn't. So, we can think about it for awhile. I think regarding 3) that I need to make POI to POI behave a little more like POI to Polygon in the strictness of name/type matching. If this example was POIs and Polys with the same attribution, I'm not sure we'd have these not so great reviews. 3) may not be terribly time consuming to make some adjustments to POI to POI to solve this problem. I'll work on that effort as part of this issue. |
I'm going to implement 2) as an option turned off by default that we can experiment with for this. Created #3579 for 3) as there's much more work required for it. |
…#3588) * Added an option, `differential.treat.reviews.as.matches`, that allows for not treating reviews as matches in Differential Conflation, which will let them pass to the diff output. The use case for this was a situation where a one to many POI review was preventing several secondary POIs from being added to the diff. Since some of these reviews were questionable, the longer term solution to this is some reworking of POI reviews (#3579). In the meantime, this config option may be useful. * Disabled the removal/replacement of roundabouts during Differential Conflation. in #3580 there was a situation where roundabouts were being mangled badly during roundabout replacement. The original point of roundabout removal/replacement was to preserve them, since we don't conflate them very well some of the time. However with diff conflate, there's no chance to mangle them since we're only keeping non-matches in the secondary data and not actually trying to merge ref and secondary roundabouts...so removing the logic was the simplest solution. * Downgraded warning logged when diff conflate w/ tags encounters a relation to a debug statement, since that is normal operation for now. #3449 can be re-opened to handle relations if need be. * Fixed a NPE in the Network alg when running ref conflate with the same data from #3387 * Fixed issue where diff conflate case tests were not actually running diff conflate * Fxied handling of output size limit in `MapComparator::_printIdDiff` * Added in some utilities to `Roundabout`, `OsmUtils`, `Node`, and `Way` that make debugging roundabout conflate problems easier
https://github.com/DigitalGlobe/VGI-team-repo/issues/1955 - Differential w/ tags and separate changeset outputs
Several POIs are being marked for review that are fairly far apart for an urban environment, have the same base restaurant type but do not have similar names or cuisines. We may be able to adjust POI to POI to prevent these reviews, which will lead to all or most of them passing through to the conflated output.
The text was updated successfully, but these errors were encountered: