Re: [talk-au] Possible mechanical edit of address names

2023-08-22 Thread Warin


On 22/8/23 14:32, Ben Ritter wrote:
Doing small targeted mechanical edits like this sounds like a good 
workflow to me.


My recommendation would be to structure the program in two steps:
1. The first script calculates all of the changes that it wants to 
make, and writes it to CSV, with all of the details required for step 2.

2. The second script reads the CSV and prepares the edit.

The primary benefit is that you can inspect the proposed edits in 
Excel and post them on a wiki page for others to see.




There are more steps than that now...

Run the render .. that generates a number of error files in text format.

Select the general error file you want to work on and put it into 
'excell' (or libreoffice cal) and convert the text to columns


Remove the errors you are not targeting this time, Get the 
nodes/ways/relations into a single sheet and save that as .csv


Now open JOSM and download the nodes/ways from that csv data .. This 
gives you the complete OSM data for the things you want to change. Save 
that .. it is an .osm file



Now open the .osm file make the changes and change the node/way/ header 
to signify that these are to be chang4ed - save that flie as the change 
file and save the one that the program is not changing to another file 
for checking and possible manual change..


Now open JOSM and open the change .osm file and download to OSM...

That is 5 steps .. checking can be done using JOSM. Looking at them from 
the csv file only gives you the osm reference number nothing else... 
you'd need to refer that number to the osm data base to see the full data.






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


Re: [talk-au] Possible mechanical edit of address names

2023-08-21 Thread Ben Ritter
Doing small targeted mechanical edits like this sounds like a good workflow
to me.

My recommendation would be to structure the program in two steps:
1. The first script calculates all of the changes that it wants to make,
and writes it to CSV, with all of the details required for step 2.
2. The second script reads the CSV and prepares the edit.

The primary benefit is that you can inspect the proposed edits in Excel and
post them on a wiki page for others to see.

It is also a sensible architecture. You can write one simple script to find
`street` -> `Street`, etc. changes and confidently submit those edits with
a quick glance. And you can also write a fancy script that uses Overpass to
find the closest similarly named street or something like that, and inspect
the proposed changes much more closely, and simply delete all the
suspicious ones. It keeps the two independent tasks insulated from each
other with the CSV. It's a perfect example of a *modular* design.

Cheers,
Ben

On Mon, 21 Aug 2023 at 16:56, Warin <61sundow...@gmail.com> wrote:

> HI
>
> I have jumped thegun here.
>
> Now looking at the data in JOSM.
>
> Total 3,814 individual addresses but
>
> The streets are only 221 so many of these are simply repartitions in the
> same street - much easier to deal with.
>
> I may simply do these manually after all. It was only the initial
> individual item numbers that frightened me. But 221 is fine.
>
>
> Note: There are more of these 'errors' I am simply picking the ones I
> can fix without resorting to the wiki for detail. Once these are done
> I'll look at the other 'errors'. While the map maker reports these
> 'errors' some of them to my human eye look to be fine... so I am not
> 'correcting' them all.
>
>
> ___
> Talk-au mailing list
> Talk-au@openstreetmap.org
> https://lists.openstreetmap.org/listinfo/talk-au
>
___
Talk-au mailing list
Talk-au@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk-au


Re: [talk-au] Possible mechanical edit of address names

2023-08-21 Thread Warin

HI

I have jumped thegun here.

Now looking at the data in JOSM.

Total 3,814 individual addresses but

The streets are only 221 so many of these are simply repartitions in the 
same street - much easier to deal with.


I may simply do these manually after all. It was only the initial 
individual item numbers that frightened me. But 221 is fine.



Note: There are more of these 'errors' I am simply picking the ones I 
can fix without resorting to the wiki for detail. Once these are done 
I'll look at the other 'errors'. While the map maker reports these 
'errors' some of them to my human eye look to be fine... so I am not 
'correcting' them all.



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


[talk-au] Possible mechanical edit of address names

2023-08-20 Thread Warin

Hi

I'm considering doing a mechanical edit of address names.

There are over 3,000 conflicts between address names and the local road 
names. The mechanical edit would target capitalization of, for example, 
'street' to 'Street'.


Source of the conflicts comes from rendering errors which are then made 
into a comma delimited list and the features downloaded by JOSM and 
saved in to a .osm file. The text to change would be limited to a list 
of words with the text starting with a space character and ending with a 
' - this removes the possibility of changing a word that contains the 
searched for text as part of an actual word.


The words to search and change would be those common such as 'street', 
'road', 'highway', 'drive', 'crescent' etc.


I'm intending to write it in Python though I am not a python programmer 
so there will be a learning curve delay.


Any thoughts particularly if you think this is a "bad idea" or some 
possible improvement.


I did write a similar program in another language, copyright so I cannot 
share it) that had echoed the change to the screen and required a 
keyboard confirmation to complete the change.. I found that to be both 
exhausting and to have found no errors. That only involved a few 
hundred, this one looks to be much larger. Of course that will be checks 
performed to ensue it does as intended.



-

I have already done some 100+ of these, a good proportion are the above 
capitalization problem. Some are things like O'riely that should be 
O'Riely .. and some are the Mc or Mac problem that does not have a set 
rule. Of course some don't fall into a category. But if 1,000 or more 
can be fixed mechanically then I'd prefer to go that way.


As for the Mc/Mac problem I'd go with the OSM road name as that is 
displayed where as the address name is hidden so the road name is more 
likely to be corrected and therefore more likely to be correct, however 
this would be done manually so is not the main topic here.



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