All, let's draw this thread to a close. I think we can safely take away that
additional training, validation, tool and process improvement is needed. Most
importantly, in the future please keep in mind -- we are all working hard
together to make the map. We owe each other respect in our communications, and
constructive comments that lead us forward. Anything else drowns out the value
of what you want to say.Thanks-Mikel
* Mikel Maron * +14152835207 @mikel s:mikelmaron
On Tuesday, April 11, 2017 9:00 AM, Ralf Stephan <[email protected]> wrote:
Just an idea. As to quality I think HOT/OSM can learn from zooniverse.org
where they have lots of projects with thousands of citizen participants that
produce science data, mostly to provide classifications for AI learning. For
example each project sporadically presents the user without telling with
pre-classified tasks in order to assess their reliability. Also, they use
classifications from several users to get the end results, that's our
validation, but can we do more like this? I'm sure zooniverse does even more
under the hood, maybe we should ask them?
Regards,
On Tue, Apr 11, 2017 at 1:53 PM john whelan <[email protected]> wrote:
Thank you Majka.
I think Majka has pinned down one major problem. Same problem as I had with
trainee programmers to them speed was important. In HOT we shoot ourselves in
the foot by saying this project is urgent, implying speed is of the essence.
We need a different way to say this.
This project is important perhaps?
Can we incorporate some elements of Majka's words into learnOSM.
Just for the record I'm not against iD I've seen someone map a building
perfectly without even touching S to square the building, but if you're mapping
buildings I don't think its the best tool for the job for new inexperienced
mappers.
The other thing that has come up on the thread is the lack of validators. Like
Polyglot I'm tired of seeing the same mistakes made over and over again. I've
cut back on validating to a single project at the moment. I've left messages
for a number of mappers only to see them make the same mistakes a week or so
later and these are mappers I've given feedback to within 24 hours of their
mapping. Yes there are some who have gone on to become solid mappers but they
aren't the majority.
At the moment I'm loading in sections of the map and correcting crossing ways,
highways that almost meet etc. normally without feedback. It cleans the map up
but it would be better if we could catch the mistakes before they are made.
JOSM will warn about crossing highways before uploading. I'm not sure if iD
etc does but there are many many many of them.
If you want more validators or people to do more validation I think you have to
ask yourselves can the job be made more attractive in some way and error
prevention might help.
Cheerio John
On 11 April 2017 at 07:25, majka <[email protected]> wrote:
I have to admit, I couldn't use iD for "bulk" mapping for the life of me. I
find it suitable for the one-off mapping / for doing corrections only. But some
mappers do and do well with this. You can find haphazardly mapped buildings and
untagged ways and nodes using JOSM for mapping as well, just not as often.A
better "building tool" for iD would help some but not for all of it.
The fundamental problem is that some mappers fail to understand mapping isn't a
race. Somehow, the number of edits / added buildings / changes became more
important than precision. We are partly promoting this by looking at the number
of edits to declare a mapper as experienced.
I try to explain to the mappers that sometimes the work is done so badly that
it would be better to do only one tenth of it but to do so correctly. As
English is my third or fourth language, I struggle with the correct way to
explain this, to get the right mix of being diplomatic and to get through -
above all when I am shouting and swearing in my head at the person who has done
the mapping.
If I could wish for one thing only to start every new mapper with, it would be
this: Exact and precise mapping is more important than anything else. Do not
map for quantity but for quality. And if unsure about tagging, look for help.
In HOT tasks, read what is expected from you and do exactly so.
Here comes the importance of earliest possible validation - to stop the bad
habit from forming. New mappers (and old ones as well) would still make
mistakes but we shouldn't ignore the systematic ones just because it is a new
mapper and we don't want to be too hard on them.
Everything else comes with experience, the speed of work as well. It is not a
problem of HOT alone - locally, a new mapper without any experience has
uploaded more than 100 changesets within the first 24 hours after his
registration. Every single one of it has to be corrected somehow. Leave it long
enough uncorrected and the map quality will degrade with tons of useless data
obscuring the correct ones.
We should somehow try to promote the idea that mapping isn't a speed race.
There are not that many tasks really time critical and even then the real
usefulness of tasks mapped just for speed is somewhat suspect. I would say, as
there are more mappers available than validators, I cannot see any reason for
"wasting" validator's time on remapping tasks. And remapping is what happens
often when validation isn't done as soon as possible and fundamental mistakes
are not caught early. I am often commenting on half finished tasks for this
reason as well - no reason to leave the problems untouched until validation.
Majka
_______________________________________________
HOT mailing list
[email protected]
https://lists.openstreetmap.org/listinfo/hot
_______________________________________________
HOT mailing list
[email protected]
https://lists.openstreetmap.org/listinfo/hot
_______________________________________________
HOT mailing list
[email protected]
https://lists.openstreetmap.org/listinfo/hot
_______________________________________________
HOT mailing list
[email protected]
https://lists.openstreetmap.org/listinfo/hot
