[HOT] Validation queries

Severin MENARD severin.menard at gmail.com
Mon Jan 20 00:19:11 UTC 2014


Hi Nick,

Sorry for the delay of my answer.

Date: Fri, 17 Jan 2014 02:14:42 +0000
> From: Nick Allen <nick.allen.54 at gmail.com>
> To: hot at openstreetmap.org
> Subject: [HOT] re  Validation queries
> Message-ID: <52D89212.5050109 at gmail.com>
> Content-Type: text/plain; charset="iso-8859-1"; Format="flowed"
>
> Severin,
>
> Not sure what I did with your original message - sorry, probably starting
> a new thread on the same topic, but this was the only way I could manage.
>
> Thank you for your detailed reply. It would indeed make a very good wiki
> article, and I would like to see that created soon, but possibly amended
> slightly.
>
Will do it soon

>
> I do have a few other thoughts about the validation process though. My
> thoughts are from my own personal experience, and it would be good to
> obtain the views of several people, preferably fairly new to the system;
>
This is defintely something I want to do

>
> When I first started mapping for HOT I found I was checking the squares
> that I had marked as complete on a regular basis. I was looking for some
> kind of confirmation that the work I was submitting had been checked
> and found to be of some kind of use. As an experienced 'mapper' I know
> that a lot of work I submit will be amended/refined in some way in the
> future by another mapper, but I felt that the validation which was
> obviously going to take place would provide some acknowledgement that I
> was producing something that moved the project forwards & wasn't hindering.
> Having looked back through the tasks I have been involved in, I can
> see that very few squares are actually marked as validated.
>
This is the issue we have and that we are trying to solve because
validation process is rarely done indeed. Currently I am running this
Activation alone and this requires a lot of time.

>
> Life experience has taught me there are circumstances where people can
> make quite serious mistakes by accident, or indeed deliberately sabotage a
> project, and I became concerned that this may not be discovered very
> quickly.
>
> Before I retired I spent many years working as a supervisor in an
> emergency control room, and found that the operators I was working with
> were keen to improve their capabilities, and prompt feedback &
> encouragement was
> extremely valuable for many reasons. In this work environment, the
> agreement between the supervisors was that we did not validate our own
> work, but that it was validated by another - thus avoiding as much human
> error as
> possible.
>
Yes, I noticed this, eg I am supporting a GISCOrps team to map Malakal in
South Sudan, answering their technical issues and monitoring theie edits
and providing feedbacks. Here again, it is a matter of time and people
available to provide this kind of support

>
> Obviously providing individual tailored feedback on each 'square' would be
> practically impossible, being extremely time consuming, and would probably
> cause offence amongst mappers who are experienced & devoted to HOT.
>
> Would it be possible, when a square is marked as validated, for some kind
> of auto generated message to be sent to the person who marked the square as
> complete - simple message something like
> /"Thank you for the time & effort you have provided in mapping square xxxx
> on HOT task 326. This square has now been validated by another experience
> mapper. If there is any particular feedback it will be listed here:-
>
>   //
> //Your further involvement in mapping//areas in which relief organisations
> are working would be appreciated & welcome.//
> //
> //Further information about the work of HOT can be found at
> .............., and if you have not already done so, subscribing to
> ........... will give you more information & updates.//
> //
> //Thanks again//
> //
> //.............. (Validator)/"
>
This message could be sent indeed, but the current validation process
through the Tasking Manager is not satisfactory for the moment, as it
allows anyone to do it, whatever his validation criteria are + it cannot be
unvalidated anymore, except by contacting Pierre Giraud. I really think
that the validation process must be done at a larger scale than the TM
tiles. We could add a check by mapper in the validation process, I will try
this. We could create a kind of matrix with quality checks, for them to
know where they could improve. It would allow us to identify good mappers,
enough to become validators. This is the requirement for any better
interaction than as for now.

Otherwise, is there a town in CAR you would like to validate?

>
> The aim of this initial validation is to encourage new mappers to HOT to
> continue, and for preference should take place fairly soon after they have
> marked the square as complete. I accept that 'Fairly soon' can be a
> difficult/impossible target when vast quantities of mappers suddenly start
> work because a disaster is imminent or has just occurred, but feel that
> encouraging new mappers to continue should provide dividends in the long
> run.
> The option to not send the message should also be available, especially in
> the event of a long delay between mapping and validation.
>
> To be feasible would require that a number of people are willing to
> validate for part of the time they devote to HOT. It would be good to
> target the HOT tasks that have involved recent activity, such as those on
> the featured
> list, but care would be needed as some mapping took place when only poorer
> or older satellite imagery was available.
>
> In the event that information about who actually marked a square as
> complete is not available, perhaps the message could be sent to the
> individual(s) who have recently mapped in that square? Just marking the
> square as
> validated would provide enough encouragement for most people who continue
> to map with HOT, but the message gives you the opportunity to pass out
> substantially more information - not sure how to manage the communication
> with different languages though!
>
> I get a great deal of satisfaction from mapping, and would not want to
> validate all the time, but feel that using it as a tool to encourage others
> to continue aiding the project is worth trying.
>
> If the HOT members feel there is some merit in trying this, I feel it
> should be documented on the wiki, so that messages do not come as a
> surprise, or worry people because they think they are being singled out.
>
> Thanks for reading - sorry there is so much of my message!
>
> Regards
>
> Nick (Tallguy)
>
> ------------------------------------------------------------------------
>
>
>   Thank you for your email. My answers inline.
>
>
> >/  Date: Sun, 12 Jan 2014 13:48:00 +0000
> />/  From: Nick Allen <nick.allen.54 at gmail.com  <
> https://lists.openstreetmap.org/listinfo/hot>>
> />/  To:HOT at openstreetmap.org  <
> https://lists.openstreetmap.org/listinfo/hot>
> />/  Subject: [HOT] Validation queries
> />/  Message-ID: <52D29D10.6010009 at gmail.com  <
> https://lists.openstreetmap.org/listinfo/hot>>
> />/  Content-Type: text/plain; charset="iso-8859-1"; Format="flowed"
> />/
> />/  Hi,
> />/
> />/  Would you like me to join in with the validation process? I am
> />/  experienced in mapping OSM, but am fairly new to regularly mapping HOT
> />/  projects.
> />/
> /Would be great! Thanks for the proposition! Indeed you have contributed a
> lot in OSM. Mapping HOT projects is not very complicated, as you saw with
> the Highway_Tag_Africa<
> https://wiki.openstreetmap.org/wiki/Highway_Tag_Africa>,
> it is less detailed then in developed countries, at least regarding remote
> mapping. Maybe the difficulty is when you do not know how those contexts
> look like. A goo way to compensate this is to look for videos posted on
> Youtube (examples
> here<http://www.youtube.com/results?search_query=bangui&sm=3>),
> especially the ones taken along road or streets. This is how you figure out
> if properties enclosures are walls, fences or hedges, what is often a
> cultural feature. Ah, just saw you mapped some wall enclosures (eg
> here<http://www.openstreetmap.org/node/27564988#map=19/4.41952/18.51870>).
> They actually are buildings (houses) under construction. It is frequent in
> developing countries that such works last a long time or even be abandoned.
>
> >/
> />/  I'm responsible for some of the mapping in
> />/  http://tasks.hotosm.org/job/72,  as well as trying to change some of
> the
> />/  more obvious 'highway=track to highway=residential or unclassified
> etc..
> />/  or it may be easier to check what I've done using my OSM profile
> />/
> />/
> http://www.openstreetmap.org/user/Tallguy/history#map=13/4.4168/18.4936&layers=N
> />/
> />/  How much validation is actually needed / done? Is it a proportion of
> the
> />/  whole task, or just until you are confident that, all things
> considered,
> />/  the task is fulfilled? You're never going to get 100% as some things
> />/  boil down to opinion about what the images actually are of, but the
> vast
> />/  proportion is pretty obvious.
> />/
> /This is something that still needs to be settled and documented. I would
> say a validation is about both identifying mistakes/mapping lacks and
> standarzation/consolidation and has 2 or three steps, related to scale:
> 1. At neighborhood scale, check notably if:
>
>     - buildings are missing. Sometimes it happens and if actually it
>     represents a consequent number of building over a TM task, it can be
>     invalidated
>     - buildings are correctly traced. Hopefully it is not frequent, but
>     sometimes mappers made really coarse outlines that do not respect
> either
>     the buildings proportions or angle. More frequent are mappers that do
> not
>     know how to square the buildings. In this case, after having checked
> what
>     is their preferred editor, I generally send a message to their OSM
> message
>     box to give them the tip to do it
>     - highway tags are correct. This is what you described. Some mappers
> put
>     tracks wherever it is not a main road considering it is not paved, but
> this
>     is not a meaningful criteria in these developing countries considering
> 99%
>     of roads are unpaved.
>     - road geometry. Some mappers do not put enough details and other too
>     much (eg a node every 10 or 20 m even if the road is straight). First
> case
>     is quickly corrected with the (magical) Improve Way Accuracy mode in
> JOSM;
>     second case requires deleting extra nodes when they actually make
> weave a
>     straight road.
>     - start/end of roads. Some mappers are experts of giant snake roads or
>     loop roads, Requires to pass the mouse over the streets to see their
> extent
>     and cut them where it makes sense. On the contrary, some streets or
> roads
>     are sawed without any reason (same tags for all the sections)
>     - general issues of connections between objects. Some that should be
>     connected and those that should not. Requires both Validator and also
> eye
>     control
>
>   2. at the town or city scale, it is quite related to the road network and
> its main highways. Having a larger view  to identify the highways that are
> not simple residential roads. They are often larger and frame a larger area
> or can be a parallel way to main roads. It is also important to check where
> they start and when they stop, what is often not possible to do when you
> map with the Tasking Manager. This is what I tried to do with
> Bouar<http://www.openstreetmap.org/#map=14/5.9374/15.5879>;
> here are some examples of issues for Bangui:
>
>     - missing parts of
> highway<http://www.openstreetmap.org/node/27564988#map=17/4.38964/18.54186
> >.
>     Looks weird on Mapnik and the check of the imagery confirms the two
> sides
>     of the road separated by a drain are not finished
>     - road continuity<
> http://www.openstreetmap.org/node/27564988#map=17/4.38686/18.50796>.
>     The situation here seems weird as well as we expect the 2 unclassified
>     roads to be connected and not joined by a simple path. The imagery
> confirms
>     that the southern highway looks the same, and should be tagged the
> same,
>     whatever the tag. The example is actually good as farther
> south<http://www.openstreetmap.org/node/27564988#map=17/4.37643/18.50931>,
>     it changes for tertiary. No reason for this, says the imagery. It
> should be
>     cut when it becomes a straight road, though
>     - isolated upper-level road
> sections<http://www.openstreetmap.org/node/27564988#map=15/4.4070/18.5068
> >.
>     Drivable highways cannot be isolated and connected to the drivable road
>     network by paths, they must be connected to it
>     - tagging coherence <
> http://www.openstreetmap.org/#map=15/4.4333/18.5314>.
>     in this example, the primary and tertiary road are connected by a
>     unclassified road. Same thing between the two tertiary roads. After
> having
>     checked the imagery this road would deserve to be tagged as tertiary.
> This
>     obvious example apart, this requires to check the streets width to
> identify
>     the main ones that needs not to be tagged as residential but
> unclassified
>     or even tertiary.
>
> Hope this can help! I had in mind to give some tips and it became a start
> for a future wikipage :) Hope other people will read/discuss/complete this.
>
> Sincerely,
>
>
> Severin
>
> Any constructive feedback from experienced HOT mappers is welcomed.
> >/
> />/
> />/  Regards
> />/
> />/  Nick/
>
> --
>
> Nick
>
> Volunteer 'Tallguy' for
> https://wiki.openstreetmap.org/wiki/Humanitarian_OSM_Team
>
> Mapping volunteer 'Tallguy' for http://www.openstreetmap.org
> <http://www.openstreetmap.org/>
>
> Treasurer, website & Bonus Ball admin for
> http://www.6thswanleyscouts.org.uk/ (treasurer at 6thswanleyscouts.org.uk
> <mailto:treasurer at 6thswanleyscouts.org.uk>)
>
> -------------- next part --------------
> An HTML attachment was scrubbed...
> URL: <
> http://lists.openstreetmap.org/pipermail/hot/attachments/20140117/825d009a/attachment.html
> >
>
> ------------------------------
>
> _______________________________________________
> HOT mailing list
> HOT at openstreetmap.org
> https://lists.openstreetmap.org/listinfo/hot
>
>
> End of HOT Digest, Vol 47, Issue 18
> ***********************************
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/hot/attachments/20140120/3130ec15/attachment-0001.html>


More information about the HOT mailing list