I wonder if it would be possible to "validate" in real time. Some method that a
validator could watch as the mapper did their thing, and make suggestions as
they went? Then it becomes both training and validation simultaneously.
Obviously it would have to be on a consent basis from the mapper, ie they would
have to opt in when they started and a validator would have to be available.
Otherwise, it seems like it would be better just to make the correction and
maybe have it send a message to the mapper about what got changed.Bryan Sayer
-------- Original message --------
From: john whelan <[email protected]>
Date: 04/24/2020 9:43 AM (GMT-05:00)
To: Jean-Marc Liotier <[email protected]>
Cc: "HOT@OSM (Humanitarian OpenStreetMap Team)" <[email protected]>
Subject: Re: [HOT] Validators [was: COVID-19 For The Long Haul]
What I found when validating HOT tasks was that if I provided feedback that
wasn't too negative on the task "ie the african highway wiki suggests that this
type of highway should be mapped thus: with a link and reference to the wiki
rather than you done it wrong" then the mapper would go on to do far more tiles
and do them correctly, but it had to be done very shortly after the mapper had
mapped. So I'd sit on a project and immediately validate the tiles as they
were done. It took considerable time and effort on my part to do this. I also
noticed some validation was done by inexperienced mappers who were invalidating
tiles for correctly mapped highways etc. One that comes to mind was project
manager who invalidated a tile for a highway that had been correctly mapped in
OSM before the project started on the grounds that mapping highways was not
part of the project.However what I noticed as Jean-Marc has commented was that
much of the mapping problems were from mappers who hadn't completed the tile
and their feedback wasn't immediate. Feedback sent to them was basically
ignored 99% of the time.A couple of things happened. The task manager changed
and I could no longer find the last tile mapped. Response tails off over time.
Within the hour 80% response, within a day 50% response, a week 2% response.
Validating a tile a week later meant the feedback didn't reach the mapper
immediately. Second I'd no idea if the mapper had corrected their problems
later or not. So the mapper might get half a dozen messages for tiles they had
mapped sometime ago and they had already changed their mapping practices. This
is not positive feedback. The second thing was the projects wanted buildings.
I'd already noticed that some buildings were being mapped two or three times.
Validating a building? Well it takes me 2 or 3 mouse clicks to map one in JOSM
with the building tool, building=yes. iD takes more clicks. To correct an
incorrect building takes more effort than to map it correctly in the first
place. Someone was kind enough to build me a script in JOSM that detects
duplicate buildings. I'm not going to validate work when it takes longer to
validate than to do it correctly in the first place.Then it gets interesting.
The JOSM validation tools have improved. These days I can grab about a tenth
of an African country at a time. It takes 16 gigs of memory and JOSM has to be
set up correctly but it works. The duplicate building script now will detect
duplicate buildings over one tenth of the country not just a tiny tile. JOSM
validation and the todo list are very powerful for picking up crossing
highways, other errors and doubtful tagging.These days there are more automated
approaches to picking up errors in mapping. Both Jean-Marc and myself fix
multiple errors by HOT mappers. I must confess that if the error was made four
years ago on a HOT project by a mapper who has made four edits and the mapper
hasn't mapped for four years I might even correct the error without a changeset
comment.I worked with the training group to identify the most common errors, I
understand that training for project managers and validators has improved and
projects no longer ask new mappers for complex mapping.I understand that
validation can help new mappers and get projects completed more quickly but I
also understand it takes a lot of time and effort on behalf of the validators
to do it right.Cheerio JohnOn Fri, 24 Apr 2020 at 07:06, Jean-Marc Liotier
<[email protected]> wrote:On 4/24/20 12:26 PM, Russell Deffner wrote:
> Especially crucial are our Validators! With so many projects in so
> many locations, it will be a monumental task to keep up on validation.
> If you have the role, please leave the mapping to the newer
> contributors and help us keep our quality up and give instructions to
> mappers as soon as possible.
What happens when tasks are invalidated ? In theory the errant
contributor welcomes constructive feedback, fixes his contribution and
mends his ways. How often does that happen in practice ? I almost (there
are a few happy exceptions of growing contributors) never receive
answers to changeset comments, so most often what was supposed to be
validation ends up as mopping up after ephemeral contributors. I don't
work with the tasking manager - my validation activities are driven by
Osmose and OSMcha... Is the validation experience different with the
tasking manager or is my experience representative of a problem ?
_______________________________________________
HOT mailing list
[email protected]
https://lists.openstreetmap.org/listinfo/hot
_______________________________________________
HOT mailing list
[email protected]
https://lists.openstreetmap.org/listinfo/hot
