Re: [OSM-talk-be] Another Nominatim fact

[André Pirard]

On 2016-09-29 08:58, joost schouppe wrote:
> Well, I was the only one thinking that about your previous mail :)
> But André is probably right: it is a strange idea to define just one
> language for a nation. I wouldn't be surprised if half the people on
> this planet live in countries with more than one official language.
> Official languages tend to follow admin boundaries, so I don't see the
> point of boundary=linguistic. But you might feed a tool like Nominatim
> with an official language tag at different admin_levels.
The point is that in Belgium, the official languages (communities) are
delimited by boundary=political boundaries in addition to the
boundary=administrative ones.
Adding tags to the administrative kind to indicate the same thing as the
political kind seems weird.
The fact is that the /political/ kind was chosen for no particular
reason (1), just because the name exists.  I wonder if it's used for
languages anywhere else than in Belgium. And hence, very few people
understand it, even on this list.
And Nominatim certainly does not.
Replacing boundary=political with boundary=linguistic would suddenly
make the issue clear for everybody.
It's like making separate boundaries for national parks and that does
not assume that they follow administrative ones (everywhere in the world).

Cheers

André.


(1) as often the case, OSM does not clearly define its words: what
"political" means, except "areas, mostly political", the opposite of
actual tagging and "electoral (?)".
> Might also help with interpreting special naming styles, like in
> Brussels. So you could have something like official_language="see
> admin level 4" for Belgium, and official_language=fr;nl for Brussels.
> Mapped this way, it might help a tool understand that you can't know
> the language of a name in Belgium by just looking at the country
> outline, and that in Brussels you have to look out for bilingual names. 
>
> As we're talking Nominatim special cases: makes me think of Philippe's
> long irritation that all addresses in central Brussels being returned
> as in the Marollen. This is because there are only thee neighborhoods
> mapped in Brussels.
>
> Proposed solution: the statistical sectors of Belgium are now open
> data. We could map the centroids of those as neighborhood nodes.
> Statistical sectors of course don't always reflect what we would call
> neighborhoods (especially in the countryside, or in industrial areas
> etc). But in city centers they do tend to reflect what might be used
> locally.
> (I don't consider this a huge problem myself, so won't work on it myself)
>
> 2016-09-29 6:47 GMT+02:00 André Pirard  >:
>
> Hi,
>
> That is exactly what I explained several times before (Nominatim's
> behavior, not that feature).
> I put it that what that feature does is: if name:ll=* is missing
> produce an implicit one with the same value as name=*.
> This means (assuming that name=* always exists), that a browser
> configured with ll as one of its primary languages will always
> find a name in language ll in a region where that language is
> primarily spoken and is Nominatim's such "default".
> Now, what you say about Flemish nl s also true for Walloon fr and
> our eastern quiet and gentle friends' de.
> So that if Nominatim is defining a language default "by country"
> as you say, they really missed something.
> They missed Belgium, they missed Switzerland, they missed Wales,
> they missed the Spanish speaking South USA etc.
>
> I have long thought of proposing a boundary=linguistic that would
> be used, typically for the Belgian regions, in parallel with the
> administrative ones and that would obviously be where Nominatim
> should pick that "default" language.
> But I have also long abandoned the idea of feeding OSM with well
> thought out suggestions because, instead of trying to understand
> my goals and possibly suggesting alternatives, my fellow
> contributors answer that this is not the way "we" do it, or other
> denials, or that I'm out of topic or even that I'm accusing people
> to "do bad job".
>
> Let it be, as the Beatles said.
> Cheers
>
> André.
>
>
> On 2016-09-29 05:20, Marc Gemis wrote:
>> Here's another fact about Nominatim that I learned after a private
>> conversation with Sarah.
>>
>> Nominatim has the possibility to install a default language for a
>> country. This is not done for Belgium, but can be done for The
>> Netherlands. Right now, the list is not complete and The Netherlands
>> is missing.
>>
>> What is the result ? In case you configure multiple languages in your
>> browser, e.g. NL & DE, you will see DE results in case there is a
>> DE-name and no explicit NL-name. This is the case for Zutphen.
>>
>> What will be the impact for Belgium ? Suppose a Flemish town is mapped
>>   

Re: [OSM-talk-be] Another Nominatim fact

[Marc Gemis]

Nominatim adds data that's not coming from OSMl. The language list I
mentioned or Tiger address data to name a few.

Maybe the problem is not that big (and that's why it's not in the
issue tracker). It will not appear if you only have 1 language
preference in your browser.
Or when the place is mapped with all name:XX tags that are applicable.
So I think you will not encounter this problem in Brussels, as all
villages etc. are mapped with name:nl and name:fr.
And if we are not creating an issue now, it might never become an
issue for Nominatim.

Your solution is better than adding data to just Nominatim, as it
allows any data consumer to reply with the correct name. Still the
data consumer needs to be updated.



regards

m


On Thu, Sep 29, 2016 at 9:58 AM, joost schouppe
 wrote:
>> As for your and André's suggestions, feel free to grab a copy of Nominatim
>> and alter it, so it recognises default languages on other admin levels.
>
>
> I was actually thinking about how to map this situation, not so much a tool
> improvement. You can hardly ask the programmers to build something if
> there's no data to feed it.
>
> That said, I thought it is allowed to make suggestions, even if you're not a
> programmer. It's just not allowed to whine about them if they aren't
> followed.
>
> A quick scan of the issues tracker of nominatim shows no sign of discussion
> about checking language at a below-national level, so they might actually
> just have overlooked it.
>
> --
> Joost @
> Openstreetmap | Twitter | LinkedIn | Meetup
>
> ___
> Talk-be mailing list
> Talk-be@openstreetmap.org
> https://lists.openstreetmap.org/listinfo/talk-be
>

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

Re: [OSM-talk-be] Another Nominatim fact

[joost schouppe]

>
> As for your and André's suggestions, feel free to grab a copy of Nominatim
> and alter it, so it recognises default languages on other admin levels.
>

I was actually thinking about how to map this situation, not so much a tool
improvement. You can hardly ask the programmers to build something if
there's no data to feed it.

That said, I thought it is allowed to make suggestions, even if you're not
a programmer. It's just not allowed to whine about them if they aren't
followed.

A quick scan of the issues tracker of nominatim shows no sign of discussion
about checking language at a below-national level, so they might actually
just have overlooked it.

-- 
Joost @
Openstreetmap  |
Twitter  | LinkedIn
 | Meetup

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

Re: [OSM-talk-be] Another Nominatim fact

[joost schouppe]

Well, I was the only one thinking that about your previous mail :)
But André is probably right: it is a strange idea to define just one
language for a nation. I wouldn't be surprised if half the people on this
planet live in countries with more than one official language. Official
languages tend to follow admin boundaries, so I don't see the point of
boundary=linguistic. But you might feed a tool like Nominatim with an
official language tag at different admin_levels. Might also help with
interpreting special naming styles, like in Brussels. So you could have
something like official_language="see admin level 4" for Belgium, and
official_language=fr;nl for Brussels. Mapped this way, it might help a tool
understand that you can't know the language of a name in Belgium by just
looking at the country outline, and that in Brussels you have to look out
for bilingual names.

As we're talking Nominatim special cases: makes me think of Philippe's long
irritation that all addresses in central Brussels being returned as in the
Marollen. This is because there are only thee neighborhoods mapped in
Brussels.

Proposed solution: the statistical sectors of Belgium are now open data. We
could map the centroids of those as neighborhood nodes. Statistical sectors
of course don't always reflect what we would call neighborhoods (especially
in the countryside, or in industrial areas etc). But in city centers they
do tend to reflect what might be used locally.
(I don't consider this a huge problem myself, so won't work on it myself)

2016-09-29 6:47 GMT+02:00 André Pirard :

> Hi,
>
> That is exactly what I explained several times before (Nominatim's
> behavior, not that feature).
> I put it that what that feature does is: if name:ll=* is missing produce
> an implicit one with the same value as name=*.
> This means (assuming that name=* always exists), that a browser configured
> with ll as one of its primary languages will always find a name in language
> ll in a region where that language is primarily spoken and is Nominatim's
> such "default".
> Now, what you say about Flemish nl s also true for Walloon fr and our
> eastern quiet and gentle friends' de.
> So that if Nominatim is defining a language default "by country" as you
> say, they really missed something.
> They missed Belgium, they missed Switzerland, they missed Wales, they
> missed the Spanish speaking South USA etc.
>
> I have long thought of proposing a boundary=linguistic that would be used,
> typically for the Belgian regions, in parallel with the administrative ones
> and that would obviously be where Nominatim should pick that "default"
> language.
> But I have also long abandoned the idea of feeding OSM with well thought
> out suggestions because, instead of trying to understand my goals and
> possibly suggesting alternatives, my fellow contributors answer that this
> is not the way "we" do it, or other denials, or that I'm out of topic or
> even that I'm accusing people to "do bad job".
>
> Let it be, as the Beatles said.
> Cheers
>
> André.
>
> On 2016-09-29 05:20, Marc Gemis wrote:
>
> Here's another fact about Nominatim that I learned after a private
> conversation with Sarah.
>
> Nominatim has the possibility to install a default language for a
> country. This is not done for Belgium, but can be done for The
> Netherlands. Right now, the list is not complete and The Netherlands
> is missing.
>
> What is the result ? In case you configure multiple languages in your
> browser, e.g. NL & DE, you will see DE results in case there is a
> DE-name and no explicit NL-name. This is the case for Zutphen.
>
> What will be the impact for Belgium ? Suppose a Flemish town is mapped
> as name=X and name:fr=Y . You install both NL and FR in your browser.
> A search will now return Y.
>
> This means we might have to map name:NL explicitly. I know some will
> consider this as mapping for the tool. Nominatim has no way (I asked)
> to do this on other levels than countries. So there is no possibility
> to tell it the default language for Flanders or Wallonia.
>
> Hope this explains why in some case you get unexpected results from
> Nominatim search
>
> regards
>
> m
>
> ___
> Talk-be mailing 
> listTalk-be@openstreetmap.orghttps://lists.openstreetmap.org/listinfo/talk-be
>
>
>
> ___
> Talk-be mailing list
> Talk-be@openstreetmap.org
> https://lists.openstreetmap.org/listinfo/talk-be
>
>


-- 
Joost @
Openstreetmap  |
Twitter  | LinkedIn
 | Meetup

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

Re: [OSM-talk-be] Another Nominatim fact

[André Pirard]

Hi,

That is exactly what I explained several times before (Nominatim's
behavior, not that feature).
I put it that what that feature does is: if name:ll=* is missing produce
an implicit one with the same value as name=*.
This means (assuming that name=* always exists), that a browser
configured with ll as one of its primary languages will always find a
name in language ll in a region where that language is primarily spoken
and is Nominatim's such "default".
Now, what you say about Flemish nl s also true for Walloon fr and our
eastern quiet and gentle friends' de.
So that if Nominatim is defining a language default "by country" as you
say, they really missed something.
They missed Belgium, they missed Switzerland, they missed Wales, they
missed the Spanish speaking South USA etc.

I have long thought of proposing a boundary=linguistic that would be
used, typically for the Belgian regions, in parallel with the
administrative ones and that would obviously be where Nominatim should
pick that "default" language.
But I have also long abandoned the idea of feeding OSM with well thought
out suggestions because, instead of trying to understand my goals and
possibly suggesting alternatives, my fellow contributors answer that
this is not the way "we" do it, or other denials, or that I'm out of
topic or even that I'm accusing people to "do bad job".

Let it be, as the Beatles said.
Cheers

André.


On 2016-09-29 05:20, Marc Gemis wrote:
> Here's another fact about Nominatim that I learned after a private
> conversation with Sarah.
>
> Nominatim has the possibility to install a default language for a
> country. This is not done for Belgium, but can be done for The
> Netherlands. Right now, the list is not complete and The Netherlands
> is missing.
>
> What is the result ? In case you configure multiple languages in your
> browser, e.g. NL & DE, you will see DE results in case there is a
> DE-name and no explicit NL-name. This is the case for Zutphen.
>
> What will be the impact for Belgium ? Suppose a Flemish town is mapped
> as name=X and name:fr=Y . You install both NL and FR in your browser.
> A search will now return Y.
>
> This means we might have to map name:NL explicitly. I know some will
> consider this as mapping for the tool. Nominatim has no way (I asked)
> to do this on other levels than countries. So there is no possibility
> to tell it the default language for Flanders or Wallonia.
>
> Hope this explains why in some case you get unexpected results from
> Nominatim search
>
> regards
>
> m
>
> ___
> Talk-be mailing list
> Talk-be@openstreetmap.org
> https://lists.openstreetmap.org/listinfo/talk-be

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

[OSM-talk-be] Another Nominatim fact

[Marc Gemis]

Here's another fact about Nominatim that I learned after a private
conversation with Sarah.

Nominatim has the possibility to install a default language for a
country. This is not done for Belgium, but can be done for The
Netherlands. Right now, the list is not complete and The Netherlands
is missing.

What is the result ? In case you configure multiple languages in your
browser, e.g. NL & DE, you will see DE results in case there is a
DE-name and no explicit NL-name. This is the case for Zutphen.

What will be the impact for Belgium ? Suppose a Flemish town is mapped
as name=X and name:fr=Y . You install both NL and FR in your browser.
A search will now return Y.

This means we might have to map name:NL explicitly. I know some will
consider this as mapping for the tool. Nominatim has no way (I asked)
to do this on other levels than countries. So there is no possibility
to tell it the default language for Flanders or Wallonia.

Hope this explains why in some case you get unexpected results from
Nominatim search

regards

m

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