Feature request: wikipedia tag normalisation

[osm-fr]

Feature request: wikipedia tag normalisation

[frodrigo]

Bonus points if this check will understand multi-language links (i.e. it will recognize fr:Canal de Suez and en:Suez Canal as same link).

This part is already present :
http://osmose.openstreetmap.fr/errors/?item=3031&class=30317

[frodrigo]

Please, can you explain the main requested feature ?

[d1g]

@frodrigo, I'm sorry for delayed answer. Take a look at http://www.openstreetmap.org/way/192205777/history and http://www.openstreetmap.org/way/135401603/history

and you will notice that exactly the same link wikipedia="ru:Ташкентский музей железнодорожной техники" was used at the same time. Usually this means that objects are duplicated or like in this case link was added twice due to editor mistake.

[d1g]

ru:Ташкентский музей железнодорожной техники and uk:Ташкентський Музей залізничної технік should be treated as it was the same object

[frodrigo]

Many parts of the same street: http://overpass-turbo.eu/?Q=way%5B%22wikiped an issue for you ?

Can point to list: http://overpass-turbo.eu/?Q=way%5B%22wikipedia%22%3D%22fr%3ACimeti%C3%A8re%20et%20m%C3%A9morial%20am%C3%A9ricain%20de%20la%20Somme%22%5D%3B%0A(._%3B%3E%3B)%3B%0Aout%3B&C=49.98592;3.22024;14&R

A brand: http://overpass-turbo.eu/?Q=way%5B%22wikipedia%22%3D%22fr%3ABurger%20King%22%5D%3B%0A(._%3B%3E%3B)%3B%0Aout%3B&C=49.06127;6.14548;8&R

SELECT
  array_agg(id),
  tags->'wikipedia'
FROM
  ways
WHERE
  tags != ''::hstore AND
  tags?'wikipedia'
GROUP BY
  tags->'wikipedia'
HAVING
  count(*) > 1
;

[d1g]

parts of the street aren't interesting, duplicated POIs are (museums, memorial, etc)

brand:wikidata=* would be a better tag in second case

[frodrigo]

I run the same on nodes/way/relation and add some filter on networks like road, rails... and other kind of brand networks. But I think there is still too many false positives.

Look on the actual output, can you suggestgest more filter to avoid false positives, please: http://sprunge.us/ZeFj

SELECT
  w,
  array_agg(tid)
FROM
(
(
SELECT
  tags->'wikipedia' AS w,
  'N' || id AS tid
FROM
  nodes
WHERE
  tags != ''::hstore AND
  tags?'wikipedia' AND
  NOT tags->'wikipedia' LIKE '%#%' AND
  NOT tags?| ARRAY['highway', 'railway', 'waterway', 'power', 'place', 'shop', 'network', 'operator']
) UNION (
SELECT
  tags->'wikipedia' AS w,
  'W' || id AS tid
FROM
  ways
WHERE
  tags != ''::hstore AND
  tags?'wikipedia' AND
  NOT tags->'wikipedia' LIKE '%#%' AND
  NOT tags?| ARRAY['highway', 'railway', 'waterway', 'power', 'place', 'shop', 'network', 'operator']
) UNION (
SELECT
  tags->'wikipedia' AS w,
  'R' || id AS tid
FROM
  relations
WHERE
  tags != ''::hstore AND
  tags?'wikipedia' AND
  NOT tags->'wikipedia' LIKE '%#%' AND
  NOT tags->'type' IN ('route', 'boundary') AND
  NOT tags?| ARRAY['highway', 'railway', 'waterway', 'power', 'place', 'shop', 'network', 'operator']
)
) AS t
GROUP BY
  w
HAVING
  count(*) > 1
;

[d1g]

• http://www.openstreetmap.org/node/4280131918
• http://www.openstreetmap.org/node/4728484984

operator=* or relation:site were used in some places to group the same operator (a website and operator name should change in one place) each office can have it's own tel. tag

operator:wikipedia or operator:wikidata would be more accurate IMO

[frodrigo]

For brands and operators I don't see what we should have ref to wikipedia. Company are abstract concept not directly linked to a place of one of his office. Maybe valid for the headquarter.

[d1g]

Company are abstract concept not directly linked to a place

thats why we are not placing wikipedia links for every mcdonalds (we have under 150 values in wikipedia=* tag)

Only headquarter of MAIF should get a wikipedia=* tag.

[didier2020]

analyser_osmosis_wikipedia use item 1220 for waterway
see http://osmose.openstreetmap.fr/en/errors/?source=29115