Re: [OSM-talk] Help with way forward with Grab

[Mateusz Konieczny]

Jan 7, 2019, 3:23 PM by mish...@mishari.net:

> 1. What's a fair expectation from such a mapping team? 
>
To not make situation worse - it is unnaceptable to make much more mistakes
than unpaid hobby mappers.

Maybe they unable to fulfil this requirement and contract signed with Grab while
remaining profitable - but it is not our problem.
 

> On one hand there's the potential to add much geometry to the project
>
low quality data is worse than no data


> Sadly Bicycle routing in the Grab mapped areas in Bangkok using Brouter.de is 
> now unreliable because service roads have been added without information 
> about access permission or barrier and in other cases ways have been 
> connected where no connection exists.
>
In cases where mistakes are obvious using available aerial images and made by a 
paid mapper
who was warned about such mistakes in changeset comments - I would revert 
immediately.

In cases where edit could not be done in a better way using aerial images - I 
am unsure and consult 
others.


> 2. Thailand's mapping community is small and it would be impossible for us to 
> go through all 7,000 edits made and set it right. I've identified at least 
> one of the issues (1-3) in roughly 1 out of 4 changesets made by Grab/Global 
> Logic in my neighborhood in Central Bangkok and I would be unable to give 
> feedback on any issues outside of this zone, likely the same with other 
> mappers.
>

I would assume that data elsewhere has the same quality. If it is making map 
worse I would revert
such edits.

> 2.1 Should we just ignore the errors, slowly fix the errors, or demand Grab 
> fixes it somehow? Can we take lessons from any other projects, say TIGER 
> imports?
>
Demand fixes before adding any new data and before subcontractor collects money 
and runs away.

> 2.2 What would even constitute as a fix by Grab? Should we ask them to do 
> something like fixme=verify_on_ground on the new entries so that it can be 
> ignored by the routing until verified?
>
The only use of fixme=verify_on_ground is that it would make mass deletion 
easier.
Data added should be correct, not marked as incorrect. And requirements for 
paid mappers
should be increased not lowered.

They should simply add correct data.

>  3.1 There doesn't seem to be a way in to enforce a commitment. 
>
If added data is known to be low quality it is OK to revert it without detailed 
review.

Sometimes just mentioning that mass revert is an option is enough.

___
talk mailing list
talk@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk

Re: [OSM-talk] Help with way forward with Grab

[Mateusz Konieczny]

Jan 7, 2019, 7:24 PM by o...@imagico.de:

>> 2.2 What would even constitute as a fix by Grab? Should we ask them
>> to do something like fixme=verify_on_ground on the new entries so
>> that it can be ignored by the routing until verified?
>>
>
> No - if you see a significant fraction of what they map is verifiably 
> incorrect they need to change their approach - either fixing the 
> errors, if necessary by going out there and checking things locally or 
> by removing the data that is faulty.
>
And in case that they continue to edit in way that adds more mistakes than a 
typical mapper 
or refuse to reply to changeset comments - it is perfectly fine to block their 
accounts (this requires
involving DWG) and revert all their edits.

___
talk mailing list
talk@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk

Re: [OSM-talk] Help with way forward with Grab

[Martin Koppenhoefer]

sent from a phone

> On 7. Jan 2019, at 18:53, Volker Schmidt  wrote:
> 
> Essentially the amount of work needed to obtain a good map for bicycles is 
> considerably higher than for a map that will only be used for car routing. 
> Mapping for bicycles requires more details, like minor roads that can onluy 
> be used by pedestrians or bicycles with accurate width plus mapping of all 
> obstacles, including the maxwidth of all barriers.


+1, I would also expect cyclists to be more demanding, a 5km detour because of 
a missing connection will maybe not even be noticed by a car driver, but a 
cyclist will be very upset when she sees it ;-)

Cheers, Martin 
___
talk mailing list
talk@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk

Re: [OSM-talk] Help with way forward with Grab

[Christoph Hormann]

On Monday 07 January 2019, Mishari Muqbil wrote:
>
> 1. What's a fair expectation from such a mapping team? 

You can and should expect at least the same level of diligence and 
cooperation as you can expect from an individual hobby mapper.

If an individual mapper maps things incorrectly in a systematic way and 
does not change this even after being urged to do so by fellow mappers 
they will get blocked and their edits reverted.  Same should apply for 
organizations.

What you can expect from an organization doing routine organized mapping 
in addition is in particular that (a) there are no beginners without 
training involved making beginners' mistakes but that they make sure 
all mappers involved have the qualification to perform the mapping 
tasks in good quality and (b) that the results of communication with 
the community propagate within the organization so you don't have to 
deal with every mapper involved separately when it comes to overall 
problems.

> 2. Thailand's mapping community is small and it would be impossible
> for us to go through all 7,000 edits made and set it right. I've
> identified at least one of the issues (1-3) in roughly 1 out of 4
> changesets made by Grab/Global Logic in my neighborhood in Central
> Bangkok and I would be unable to give feedback on any issues outside
> of this zone, likely the same with other mappers.

I think when you deal with large volume organized activities it is fair 
if mappers use sample tests to assess the quality of mapping and 
extrapolate to the whole.  The burden of proof that other parts of the 
mapping are better quality than the samples evaluated lies with the 
organization.

> 2.2 What would even constitute as a fix by Grab? Should we ask them
> to do something like fixme=verify_on_ground on the new entries so
> that it can be ignored by the routing until verified? 

No - if you see a significant fraction of what they map is verifiably 
incorrect they need to change their approach - either fixing the 
errors, if necessary by going out there and checking things locally or 
by removing the data that is faulty.

> 3. It's been difficult to agree with the mapping team on a standard.

It is important for you to keep in mind that you as the local hobby 
mappers are the guardians of the data quality in your area.  Ultimately 
the only real pressure you can put on organizations to ensure that is 
of course reverting their edits.  So if you find they are not able or 
willing to make sure what they map satisfies the standards of quality 
of the local community you only have two options:  Make clear that 
their edits will be reverted in bulk if they don't meet your quality 
standards (and if necessary revert them) or adjust your quality 
standards to theirs.  I know this might be a difficult choice if they 
also make a lot of useful edits.  You need to decide if that is worth 
more than the damage from the mapping errors.  But if you make clear 
the local community will revert their edits if they don't substantially 
improve their quality that will likely at least bring some movement 
into things.

Apart from that what i in general like to suggest to local communities 
dealing with organized remote activities is to put pressure on them to 
hire locals instead of people in some distant country to perform 
mapping tasks.  First this is good for your local economy and second 
people mapping their local environment have a higher inherent incentive 
and ability to map in good quality than remote mappers mapping in a 
country they don't know first hand.

What i also would suggest is asking your forum moderator to block 
collective accounts from organizations.  The OSM communication channels 
are for individual humans to communicate.  They may use pseudonyms but 
i would not accept communication with a collective entity behind which 
an arbitrary number of people might hide.

-- 
Christoph Hormann
http://www.imagico.de/

___
talk mailing list
talk@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk

Re: [OSM-talk] Help with way forward with Grab

[John Whelan]

It's a difficult call.  I do a fair amount of armchair mapping and in 
general I use the lowest classification of highways that looks 
reasonable when I add a highway,if it should be higher then I leave the 
change to the locals.  In general I do not delete or change tagging 
except when buildings have been duplicated when I'll delete the 
duplicates if I'm feeling nice.


Having said that when I see two small villages joined by what looks like 
an unconnected 2 km narrow dirt highway tagged as a motorway or primary 
highway I will change the tag to unclassified.


Many of the problems come down to training.  HOT has some very 
reasonable training guides.


At the end of the day the company commissioning the work want accurate 
maps so the goals are the same.   I'd suggest a team approach, simpler 
tasks are given to new mappers, more complex tasks such as changing tags 
or guessing if two highways are connected be given to ones with more 
experience.


Occasionally I'll see something that looks odd.  If it was done three 
years ago by  someone who hasn't mapped in three years and only made 
four edits I'll probably correct it. If it was made by an active mapper 
with a few thousand changesets to their name we'll have a discussion.  
If it was mapped by Fredrick Ramm I might even leave it untouched.


If it says "I love Elsie" with a heart shape I have been known to delete 
it without reference to the mapper.


Cheerio John

Mishari Muqbil wrote on 2019-01-07 9:23 AM:

Hi Everyone,

Some people from the Thai mapping community are meeting Ajay, the 
mapping lead from Grab to discuss the quality issues we've been having 
with their mappers but before we do this I wanted input from the wider 
community. For those who are unaware, Grab has commissioned another 
entity, Global Logic, to do the edits.


I've made some example of quality issues here: 
https://wiki.openstreetmap.org/wiki/Grab/Issues I think I need more 
help to categorize the problems and propose remedies for them 
especially in the context of directed mapping teams. There were also 
more issues documented in this thread 
https://forum.openstreetmap.org/viewtopic.php?id=6407 but I haven't 
sorted through all of them. This process would at least take the focus 
away from individual problems and allow us to come up with a strategy 
for dealing with classes of problem instead.


From what I can tell, these are the broad classes of issues:
1. Connecting roads where no connection exists - ie. making guesses 
where there are buildings or tree covers blocking satellite imagery.

2. Incorrect Tagging - i.e. residential roads in areas marked as parks
3. Not using up to date satellite imagery and overriding recent 
changes made by local mappers
4. Changing existing tags - ie. secondary_link to primary etc. 
overriding local mappers discretion.

5. Micro-alignment using satellite imagery

What I would like help with is this:

1. What's a fair expectation from such a mapping team? On one hand 
there's the potential to add much geometry to the project, on the 
other hand any bad data has an outsized effect on utility and 
perceived quality. What's the sweet spot? Sadly Bicycle routing in the 
Grab mapped areas in Bangkok using Brouter.de is now unreliable 
because service roads have been added without information about access 
permission or barrier and in other cases ways have been connected 
where no connection exists.


2. Thailand's mapping community is small and it would be impossible 
for us to go through all 7,000 edits made and set it right. I've 
identified at least one of the issues (1-3) in roughly 1 out of 4 
changesets made by Grab/Global Logic in my neighborhood in Central 
Bangkok and I would be unable to give feedback on any issues outside 
of this zone, likely the same with other mappers.
2.1 Should we just ignore the errors, slowly fix the errors, or demand 
Grab fixes it somehow? Can we take lessons from any other projects, 
say TIGER imports?
2.2 What would even constitute as a fix by Grab? Should we ask them to 
do something like fixme=verify_on_ground on the new entries so that it 
can be ignored by the routing until verified? At least it will make 
the map usable again. Alternatively we identify the sweet spot in (1) 
and ask Grab to redo their edits to comply.


3. It's been difficult to agree with the mapping team on a standard. 
They did make some statements in this post 
 
which seems to highlight a difference of understanding between the 
community and the mapping team.
3.1 There doesn't seem to be a way in to enforce a commitment. 
Hypothetically if they said "we do QC" then we continue to uncover 
errors or “we don’t retag” when clearly this is happening, we can’t 
stop them.
3.2 Maybe there needs to be a way of assessing mapping activities so 
new organized mappers know what would qualify as a high quality 
contribution.


4. The directed editing policy