Re: [Tagging] Is there a way to make tags better?

[Lorenzo "Beba" Beltrami]

2017-03-16 18:55 GMT+01:00 Janko Mihelić :

> What about using an existing solution, Wikibase (of which Wikidata is an
> example)?
>
> [...]
>

> And countless other examples...
>

+1
I was thinking about this for the wiki translations.
A tag description has to be translated, of course, in every language. But
some data, like related tags or link to proposition or many parts of
{{ValueDescription}} template[1], could be generalized sparing much of the
translation work resulting in a more updated wiki.

Lorenzo

[1] https://wiki.openstreetmap.org/wiki/Template:ValueDescription
___
Tagging mailing list
Tagging@openstreetmap.org
https://lists.openstreetmap.org/listinfo/tagging

Re: [Tagging] Is there a way to make tags better?

[Janko Mihelić]

What about using an existing solution, Wikibase[1] (of which Wikidata is an
example)? Let's call it OSMdata. Wikibase is great because it allows for
all kinds of structured data, which makes it future-proof. Some examples of
use would be:

We could add links to outside databases, for example Wikidata, or some
other structured source that would add semantic meaning to some specialized
tags. So if there is an unique ID of a special type of buoy, connect it to
our seamark:type=cardinal_buoy tag on OSMdata.

Create items that are tag combinations, and give those combinations
additional data. For example, create an OSMdata item for
amenity=place_of_worship+religion=buddhist, link it with the Wikipedia
article about Buddhist temples. This gives more info about tag
combinations, and in some way formalizes those combinations.

Translations to all possible tag combinations (usable by all editors, and
Nominatim for example)

Add additional "meta tags", depending on the region they are found in. For
example, default max_speed on highway=motorway in Spain.

Create items for big geographical entities, like continents and oceans, and
then, instead of relations, tag boundaries of those objects with the
OSMdata ID.

And countless other examples...

[1] https://www.mediawiki.org/wiki/Wikibase

Janko Mihelić
___
Tagging mailing list
Tagging@openstreetmap.org
https://lists.openstreetmap.org/listinfo/tagging

Re: [Tagging] Is there a way to make tags better?

[Imre Samu]

>Thoughts?

Unified tag system (iD, JOSM) + suggestions + sharing translations  +
unified metadata  ==   imho :  one of key problems,  but not so easy to
solve.

Some related projects :

#1. Unified Preset Management System ( Google Summer of Code/2017/Project
Ideas )

https://wiki.openstreetmap.org/wiki/Google_Summer_of_Code/2017/Project_Ideas#Unified_Preset_Management_System

#2. iD Editor has a synonyms for searching for the preset (
https://github.com/openstreetmap/iD/tree/master/data/presets )
like searching any of  ["elderly", "living", "nursing",
   "old",   "senior"]  ->  "Nursing Home"
   "tags": {
"amenity": "social_facility",
"social_facility": "nursing_home",
"social_facility:for": "senior"

https://github.com/openstreetmap/iD/blob/master/data/presets/presets/amenity/social_facility/nursing_home.json

#3 (iD) Name Suggestion Index
https://github.com/osmlab/name-suggestion-index
 and my favorite issues:
 -   https://github.com/osmlab/name-suggestion-index/issues/11
 ( Localization of name suggestion )
 -   https://github.com/osmlab/name-suggestion-index/issues/38  (
wikidata, McDonald's vs. Мекдоналдс, wikidata property (P1282)  )
imho:  the machine learning techniques to make suggestions - can be used
here ..

#4 https://github.com/geometalab/OSMTagFinder
   "A search engine for OpenStreetMap tags. It uses TagInfo, translation
services (german to english), thesaurs and an adapted domain-specific
semantic net."

> There should be a way for community to suggest tag replacements.

iD Editor has  deprecated.json metadata :
https://github.com/openstreetmap/iD/blob/master/data/deprecated.json
 ( "amenity=firepit"  -> "leisure=firepit")
maybe JOSM has a similar metadata.



2017-03-16 1:57 GMT+01:00 Yuri Astrakhan :

> TLDR: Proposing several technologies to organize tags and help new users.
>
> With the rapid community growth, the same concepts tend to be described in
> more and more ways (tags/values), making the data maintenance and
> consumption increasingly difficult. taginfo site is an amazing effort to
> bring order to this space, and I would like to discuss what we can do to
> systematically improve it even further.
>
> == Suggestions ==
> In addition to presets, I think there should be an suggester service.
> JOSM, iD and other programs can use it to offer a list of suggested
> tags/values when editing, and to highlight those tags/values that appear
> significantly out of place.
> The suggested list should be the result of a complex search based on the
> position, size, type, and all other tags and their values present on the
> given object.  We may even want to use machine learning techniques to make
> suggestions.  I am exploring if ElasticSearch can help (disclaimer: I work
> @ Elastic).
>
> == Precise Meaning ==
> Whenever someone enters a tag=value pair, they have a very specific
> meaning in mind. If that meaning is misunderstood by other mappers or data
> consumers, that tag causes more damage than no tag - it's misleading. At
> the same time, we do not want to limit new meanings from being added to
> OSM.  I think there should be a minor hinderance when adding a new tag or
> value (for enum-style tags). Whenever user tries to add a new tag, they
> should be shown a popup to provide a brief description of the new tag, and
> explain how it differs from the existing tags. The popup may also suggest
> existing tags based on the description (similar to how stackoverflow shows
> existing questions when user is typing a new one).
> This system should be enabled for tags and for "enum-like" values (tags
> like "name" clearly should not cause a popup, whereas unusual value for
> "boundary" tag should show a popup).
>
> == Cleanup suggestions ==
> There should be a way for community to suggest tag replacements. For
> example, while browsing taginfo, I saw religion=catholic and religion=
> católico, and I can suggest that it should be replaced with
> religion=christian & denomination=catholic.  Afterwards, a site
> like MapRoulette could allow users to quickly check that replacements
> actually makes sense for each object, and accept or reject it.
>
> Thoughts?
>
> ___
> Tagging mailing list
> Tagging@openstreetmap.org
> https://lists.openstreetmap.org/listinfo/tagging
>
>
___
Tagging mailing list
Tagging@openstreetmap.org
https://lists.openstreetmap.org/listinfo/tagging

Re: [Tagging] Is there a way to make tags better?

[Frederik Ramm]

Hi,

   there have been many plans/concepts in the past that tried to solve
the same problems. I have lost track of the different endeavours and
what they wanted to achieve and why they didn't. I remember a talk by
David Earl about something tagging-related at a SOTM conference years
ago; I remember some work being done by Stefan Keller at Rapperswil, and
also quite recently someone did a little "big data" stuff with tags and
tried to automatically find out "if you tag this you might also want to
use these tags" rules or so. You should definitely also google for
"Harry Wood Community Smoothness" and view that 2009 talk of his.

On 16.03.2017 01:57, Yuri Astrakhan wrote:
> == Suggestions ==
> In addition to presets, I think there should be an suggester service.
> JOSM, iD and other programs can use it 

Precursors of such a suggester service are probably the presets offered
by all big editors, and in the case of JOSM also the validator.

Currently there's not even a database of presets or validation rules
shared by all editors. Maybe that could be a first step (having a
"preset service" and a "validation service", but it would need to
provide some kind of offline capability too for editors that can't rely
working internet connection while being used).

Great power lies in creating and maintaining these services because you
can essentially control what tags are used "in the wild". If the
introduction of new tags is made more complicated, the power of the
maintainers will be even greater. This is something that we, as a
community, should think about - whom do we want to have this power?

> We may even want to use machine learning techniques to
> make suggestions.

I am critical of that because it would remove the community's ability to
check the algorithms. Currently I can look at the code of an editor and
discern in which situations it will offer which choices to the user, and
modify that if necessary. Once machine learning is employed, this will
be more difficult, and editors could start suggesting stuff that we
don't even want them to.

> == Precise Meaning ==
> Whenever someone enters a tag=value pair, they have a very specific
> meaning in mind. If that meaning is misunderstood by other mappers or
> data consumers, that tag causes more damage than no tag

I'm not sure that (a) this is true and (b) the mapper must make more of
an effort to be understood - maybe it is the consumer who must make more
of an effort to understand?

> they should be shown a popup 

There's a long-going and mostly humorous feud between JOSM-leaning
people ("let's show a popup") and ID-leaning people ("gah! popups ruin
the UX").

> == Cleanup suggestions ==
> There should be a way for community to suggest tag replacements. For
> example, while browsing taginfo, I saw religion=catholic and religion=
> católico, and I can suggest that it should be replaced with
> religion=christian & denomination=catholic.  Afterwards, a site
> like MapRoulette could allow users to quickly check that replacements
> actually makes sense for each object, and accept or reject it.

This is a very contentious topic even if you leave religion out of it.
There's a ton of things that can go wrong. In your example, one would
have to ensure that only people who understand religion in the
particular country will participate in the verification (plus there will
be "helpful" people trying to automate the process). At DWG we have to
deal with such well-meaning but erroneous mass edits all the time. For
example (that was years ago), someone in Germany tried to change all
"denomination=evangelical" to "denomination=protestant" in Germany
because - that was his reasoning - the German term for "protestant" is
"evangelisch" leading to many mappers mis-mapping a protestant church as
"evangelical". But of course some of them were "evangelical" on purpose.

Even today, some people are very eager to "deprecate" old tags in favour
of something new they think should be used instead. Giving them a
streamlined process to get rid of "old" tags sounds like inviting trouble.

Bye
Frederik

-- 
Frederik Ramm  ##  eMail frede...@remote.org  ##  N49°00'09" E008°23'33"

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

[Tagging] Is there a way to make tags better?

[Yuri Astrakhan]

TLDR: Proposing several technologies to organize tags and help new users.

With the rapid community growth, the same concepts tend to be described in
more and more ways (tags/values), making the data maintenance and
consumption increasingly difficult. taginfo site is an amazing effort to
bring order to this space, and I would like to discuss what we can do to
systematically improve it even further.

== Suggestions ==
In addition to presets, I think there should be an suggester service. JOSM,
iD and other programs can use it to offer a list of suggested tags/values
when editing, and to highlight those tags/values that appear significantly
out of place.
The suggested list should be the result of a complex search based on the
position, size, type, and all other tags and their values present on the
given object.  We may even want to use machine learning techniques to make
suggestions.  I am exploring if ElasticSearch can help (disclaimer: I work
@ Elastic).

== Precise Meaning ==
Whenever someone enters a tag=value pair, they have a very specific meaning
in mind. If that meaning is misunderstood by other mappers or data
consumers, that tag causes more damage than no tag - it's misleading. At
the same time, we do not want to limit new meanings from being added to
OSM.  I think there should be a minor hinderance when adding a new tag or
value (for enum-style tags). Whenever user tries to add a new tag, they
should be shown a popup to provide a brief description of the new tag, and
explain how it differs from the existing tags. The popup may also suggest
existing tags based on the description (similar to how stackoverflow shows
existing questions when user is typing a new one).
This system should be enabled for tags and for "enum-like" values (tags
like "name" clearly should not cause a popup, whereas unusual value for
"boundary" tag should show a popup).

== Cleanup suggestions ==
There should be a way for community to suggest tag replacements. For
example, while browsing taginfo, I saw religion=catholic and religion=
católico, and I can suggest that it should be replaced with
religion=christian & denomination=catholic.  Afterwards, a site
like MapRoulette could allow users to quickly check that replacements
actually makes sense for each object, and accept or reject it.

Thoughts?
___
Tagging mailing list
Tagging@openstreetmap.org
https://lists.openstreetmap.org/listinfo/tagging