Interesting. For your information, I've recently written a document in which I describe what I think could be the future enhancements in the tasking manager. Specifically, there's a section about validation you may want to read. https://docs.google.com/document/d/1l3PwPbUPfXptQumZK_a_xioPYiGXAAw8JpKQhADmsJk/edit?usp=sharing
Regards, Pierre On Mon, Aug 24, 2015 at 6:40 AM, Suzan Reed <su...@suzanreed.com> wrote: > Friends, > > I would like to see some boundaries set on who validates. Someone with less > than 50 changesets should not be validating, or even marking a tile Done. > Sometimes I think new mapper validate and invalidate just for the fun of it. > They quickly validate or invalidate and don’t map or complete a tile. > > Would it be possible to require a validator to !. complete the validation > course, and 2. have a set number of tiles completed? I know this will > probably be an arbitrary number and we can all argue all the many benchmarks, > but at least set it high enough so newbies are validating. > > I’ve often thought the Tasking Manager could use a little tweaking so the > Done, Validate, and Invalidate buttons are more intuitively understood. I’ve > seen newbies mark a tile Done when they finish their session. I think I might > have done that in my first days, too. > > Like John’s idea of having a designated validation person or team watching > for tiles that need validation in a specific project. Blake Girdardot > actively involved in the Myanmar mapping, and it was very helpful. > > Possibly the validation course needs a bit of editing. These modules are up > for comments. Currently a team is editing for tone and clarity, but it is not > integrated into the modules yet. > > Suzan > > > > On Aug 23, 2015, at 4:34 PM, john whelan <jwhelan0...@gmail.com> wrote: > > I've been looking through the new courses and one thing that hit me was my > idea of validation seems quite different to the concept I've seen so far in > the course so I think we should start by deciding what we want our tile > validators to do. > > These are my thoughts. > > Higher level validation can use different tools over a wider area. > > First comment is what I've found to be the most successful is not to declare > something invalid unless its really bad. You want the mapper to feel > welcome, you want them to map again, INVALID missing a hut doesn't do that. > I've had people send me that on one of my tiles by the way, just map the hut > and move on. > > Generally I'll sit on one or more projects and validate just those projects > as the tiles are done. > > The objective is to give feedback within 24 hrs or less to the mapper. > > This feedback serves two purposes, one we are interested in your mapping and > second the earlier I can catch someone making a mistake the fewer errors I'll > need to fix in the future. > > In Cameroon using these techniques we've actually managed to completely map > and validate several projects. > > Note to Project managers if you want your project completed get yourself a > validator who validates the tiles as they are done. > > When you start on a new project take a look at the mapper, if they've mapped > twice three months ago then don't waste your time sending them emails just > clean up. > > Personally I only use JOSM when validating, you do have to press the validate > button for it to do its thing by the way. It will only automatically > validate those edits you have made when you upload not the rest. > > Having said that there is a place for iD when validating, two of the mappers > I work with validate as a team one does the careful visual checking in iD, > the other runs JOSM over the end product. > > The quality of the imagery used seems to have an impact on the quality of the > mapping. Especially with new mappers, less than ideal imagery means > validation is slow and tedious. > > I think we have to ask ourselves about how much we are prepared to pay for > what quality of work. ie service level agreement. > > When we have a very large number of new mappers who are making lots of errors > then sometimes the judgement call is a JOSM validation to clean up the worst > errors and tag it "validated in JOSM" so if someone has the time they can go > back over it. Select two top tiles and two base tiles and bring them into > JOSM, now download the area between directly from OSM run the validator tool > and do the search checks, ie area=yes etc. Its fast and picks up many > mistakes but isn't the same quality as a normal validation. > > Things to look for are untagged ways - JOSM validation will pick these up. > > area=yes can be landuse=residential or building=yes > zebra crossings in anywhere but the UK shouldn't be there. > > crossing highways not connected, throws the routing software. > > In Africa in rural areas highway=footway should be highway=path same for > highway=pedestrian, careful how you give feedback if they are an experienced > OSM mapper they're used to tagging with other values and you want to retain > them so point them to the African highway wiki bit and suggest > highway=unclassified, track or path are the most commonly used values in > rural areas. > > Buildings not squared, select buildings=yes the select each mapper in turn if > the don't have any huts use q to square them all at once. This is a time > management issue if we had more time we should do them one at a time reality > does it really matter if the building is square? We know the rough shape and > size and buildings are expensive in mapper time. Locally talking to one > mapper, buildings are OK for the first three hours after that forget them. > Another mapper I know refers them as building hell projects. Some mappers > are very good and will map buildings, I tend to be protective of them and > very gentle with them as well. If you have a building project the ones > mapped in JOSM building_tool are much easier to validate than the iD ones. > > Scan the tile <crtl><Down arrow> and look for missed settlements, I think my > record is twenty on a single tile. > > huts are awkward, OSM says point or circular, HOT prefers circular but > experienced mappers will often use points. You want them to keep HOT > mapping, if you criticise them too much they'll go back and map waste paper > bins locally. > > Is there some sort of path or unclassified highway to each settlement? If > not can we drop one in. I map fewer paths, and tracks than I used to, it > takes time to map every track and path. > > Try to avoid being condescending when communicating with mappers. "The > content is great, but how things are said often leaves the impression we are > a bunch of arrogant, condescending jerks." Each mapper is different, you > won't have time to read their life history before contacting them, click on > their name the OSM profile send message or @username (@[firstname secondname] > if they have a space in their name) in the comment. the @ method works best > as it shows the project and tile numbers. Be polite and if possible use the > third person, JOSM validation threw up the following problems on this tile. > HOT may not have their email address but you can contact them, the OSM > profile will send an email to their email address. > > If they are carefully mapping every building on the project and it isn't > required in the instructions I tend to say thanks for mapping them but they > aren't required in the instructions, hoping they'll map a few more tiles more > quickly. > > Somethings are subjective, if one mapper has mapped them one way on one tile > and another mapper has mapped them differently on another unless there is a > clear way to map them I'd let them go. For example highway=unclassified or > highway=path sometimes its difficult to be definite one way or the other > especially when you look at the number of buildings in the settlement. > > I do do a <crtl>f with nothing in it to select everything then scroll down > the list of tags looking for unusual ones. > > Finally I never use JOSM Presets <crtl>f works fine for me. > > I'm sure others have different ways of validating so your thoughts please > then perhaps we can first define our target audience then refine the course. > > Cheerio John > > > > _______________________________________________ > HOT mailing list > HOT@openstreetmap.org > https://lists.openstreetmap.org/listinfo/hot > > > _______________________________________________ > HOT mailing list > HOT@openstreetmap.org > https://lists.openstreetmap.org/listinfo/hot -- ------------------------------------------------------------- | Pierre GIRAUD ------------------------------------------------------------- _______________________________________________ HOT mailing list HOT@openstreetmap.org https://lists.openstreetmap.org/listinfo/hot
