Italia aggiunta al matcher sperimentale di All The Places

Communities Italia (Italy)
1

Per chi non fosse nel gruppo Telegram italiano o si fosse perso il messaggio: da due settimane l’Italia è stata inclusa nel matcher sperimentale di matkoniecz. Questo tool è progettato per migliorare i dati relativi ai POI presenti su All the Places.

Potete raggiungerlo a questo link: LINK

Potete concentrarvi su “Name mismatch” e “Tags to add”, la categoria “Missing in OSM” e “Location mismatch” invece hanno molti falsi positivi, presumo dovuto alla mancata localizzazione “(not attempted location check for x)”.

2 Likes
2

@Mateusz_Konieczny thanks for adding Italy! Regarding:

how does the “location check” work?

1 Like
3

sembra che per l’aggiunta di tag, il sistema non noti la presenza dei contact:phone proponendo comunque il tradizionale phone

1 Like
4

Il dataset Bricofer contiene anche i Self e il sistema non e’ in grado di capire che sono gia’ presenti.

1 Like
5

it uses address tags to run Nominatim query - if location heavily mismatches what was claimed by ATP it goes into special category

though as I am using public nominatim server I am limited to 1 query/s and it is also making some other queries - it has not completed its run across all shops (and may never complete it, I am looking for solutions)

1 Like
6

Fixed, thanks for spotting. BTW, phone suggesting is disable for now as I noticed that for many entries it proposes call centers, the same across all locations.

Not sure is it welcome in phone tag and definitely mismatches promised data contents from ATP. There are plans to fix it on ATP side so I will wait a bit.

Fix is in handle contact:phone · 4722a5f03f - matkoniecz/list_how_openstreetmap_can_be_improved_with_alltheplaces_data - Codeberg.org (with regression tests!)

1 Like
7

Autotranslate gets confused :frowning:

I looked at some at ATP <-> OSM bricofer_it - missing shops? and found no clear problems. Can you point out specific case?

8

Apparently in 2018 the company “Bricofer Italia” bought the company “Self Italia”, so they now list Self shops into their website. But on the ground, those shops still have “Self” as a name, so the matcher can’t recognize them. E.g. this shop (website / OSM):

image
image452×793 88.3 KB

I wonder if adding the Bricofer brand:wikidata while leaving the Self name would make the POI disappear from the missing locations of the matcher and make it appear in the “Tags to add” category?

1 Like
9

It seems this info is not present on Negozio Bricofer di Vercelli so ATP is helpless here.

I will switch off this spider.

10

Il sito é stato recentemente aggiornato, ora non é diviso per nazioni ma in una griglia che dovrebbe permettervi di trovare match più locali, inoltre sono supportati molti più brand, dategli un’occhiata: https://matkoniecz.codeberg.page/improving_openstreetmap_using_alltheplaces_dataset/

11

@Mateusz_Konieczny not sure if I have to report this to you or to the AllThePlaces people, I noticed a couple things for the Yves Rocher brand:

image

  1. it suggest shop=beauty, in my experience they are (mainly?) shop=cosmetics uhm, seems like there’s no consensus or the are different “versions” of the shop (one with beauty services and one without them). NSI has 5 Yves Rocher entries, 1 is =cosmetics and 4 are =beauty + beauty=cosmetics. Also we have 48 Yves Rocher mapped in Italy, 24 are =beauty, 23 are =cosmetics and 1 is =chemist.
  2. it suggests to add weird logos as image=* values e.g. link
  3. is it possible to disable the addr:state=* suggestions? is not a tag we really use in Italy
  4. this is to report to the NSI people I think but I write it here just for the record: the matcher suggest brand:wikidata=Q1477321, but iD for the same brand suggests Q28496595, which is not the brand wikidata but the company-that-owns-the-brand’s wikidata.

image
image1391×127 8.84 KB

I’m curious, why is Tesla’s percentage at 67% if all the columns are at 0?

12

I am reporting entries as matching if close to POI and as missing when far away.

When in middle it goes onto “Uh, maybe” pile that is neither shown (as many actually have matches) nor considered as matching.

Yes, outcome is confusing and should be improved, not sure about best solution.

the tricky part is that these make sense in some places

maybe I should check addr:state ? And keep it only for USA?

I will report it to ATP - done at yves_rocher_it has weird logos as images · Issue #10595 · alltheplaces/alltheplaces · GitHub

13

Also, I created more systematic report about bogus images and several more spiders were fixed - see cases where image is the same in over 500 POI, indicating that it is likely not POI-specific · Issue #10630 · alltheplaces/alltheplaces · GitHub

And next one with more of them: cases where image is the same is repeated - more than 20 times, less than 500 times · Issue #10679 · alltheplaces/alltheplaces · GitHub

Also, no matter whether ATP will fix it or not, these images will be discarded by my tool, see discard bad images · bb5418f582 - matkoniecz/list_how_openstreetmap_can_be_improved_with_alltheplaces_data - Codeberg.org

1 Like