Can't find address when using блок instead of бл.

[Dimitar5555]

What did you search for?

блок 717 ж.к. люлин 7
https://www.openstreetmap.org/search?query=%D0%B1%D0%BB%D0%BE%D0%BA%20717%20%D0%B6.%D0%BA.%20%D0%BB%D1%8E%D0%BB%D0%B8%D0%BD%207

What result did you get?

Nothing

What result did you expect?

The same as when searching for бл. 717 ж.к. люлин 7
https://www.openstreetmap.org/search?query=%D0%B1%D0%BB.%20717%20%D0%B6.%D0%BA.%20%D0%BB%D1%8E%D0%BB%D0%B8%D0%BD%207

When the result missing completely:
https://www.openstreetmap.org/way/26667926

Further details

блок is the long version of бл.. It is also listed in https://github.com/osm-search/Nominatim/blob/master/settings/icu-rules/variants-bg.yaml.

[lonvia]

Nominatim only abbreviates names coming from OSM, not the other way around because it is a long-standing rule that OSM should not contain abbreviations. Looks like Bulgaria is mapped in violation of this rule at a large scale.

[Dimitar5555]

Bulgaria is indeed mapped using a lot of abbreviations but it's because we don't have cases where a single abbreviation can be expanded to 2 or more words with different meanings (like in English St can be expanded to Street or Saint).

[lonvia]

OSM is a global project. The abbreviations can extend to different words in other languages.

[Dimitar5555]

I've changed addr:housenumber's value to блок 717. Both бл. 717, ж.к. люлин 7 and блок 717, ж.к. Люлин 7 do not work now. Am I missing something?

[Dimitar5555]

Nominatim only abbreviates names coming from OSM, not the other way around because it is a long-standing rule that OSM should not contain abbreviations.

It seems like the wiki has conflicting information on this topic. See https://wiki.openstreetmap.org/wiki/Name_finder:Abbreviations

Abbreviations work both ways: so if someone has tagged a street Saint Johns St, then users can also expect a match for St Johns St, St Johns Street and Saint Johns Street (and also, incidentally, St John's St or St John St etc.), and vice-versa.

[lonvia]

That's because Nominatim penalizes housenumbers with lots of letters in them up to a point where it cannot find them anymore. This should get fixed with the modified search algorithm currently in the works for the Python implementation.

The underlying problem is that addr:housenumber contains something that is not really a housenumber. If I understand the system right, then бл. 717, ж.к. люлин 7 isn't even a full address yet. It is missing at least the entrance number for the block. So a system where the apartment building simply gets a name=блок 717 and the entrances something like addr:housenumber=1+addr:block=717+addr:... might be better to capture the addressing system. But I'm speculating here. addr:block doesn't really have a standardized use in OSM and is not yet implemented in Nominatim.

[Dimitar5555]

This should get fixed with the modified search algorithm currently in the works for the Python implementation.

Is there an ETA for it?

The underlying problem is that addr:housenumber contains something that is not really a housenumber. If I understand the system right, then бл. 717, ж.к. люлин 7 isn't even a full address yet. It is missing at least the entrance number for the block. So a system where the apartment building simply gets a name=блок 717 and the entrances something like addr:housenumber=1+addr:block=717+addr:... might be better to capture the addressing system. But I'm speculating here. addr:block doesn't really have a standardized use in OSM and is not yet implemented in Nominatim.

addr:block seems to be used mainly in India as a replacement of addr:place (or at least so it seems).

For the entrances we use entrance=staircase + ref=* without duplicating addr:* information since it can be automatically copied from the building footprint.

addr:housenumber represents the building number within that neighborhood (the neighborhood acts like addr:street). блок is used do distinguish between house numbers and apartment buildings. It's not possible to remove it since there are places where on a single street you can find both addressing systems. Of course, the numbers will match so you wouldn't know which of the two numbers 4 is the house and which is the apartment building.

Nominatim only abbreviates names coming from OSM, not the other way around because it is a long-standing rule that OSM should not contain abbreviations. Looks like Bulgaria is mapped in violation of this rule at a large scale.

Is it possible for Nominatim to perform a second search after the regular one but with abbreviated search terms or that would add too much complexity?

Another thing to note about abbreviations in Bulgaria is that there are almost no modern paper/digital Bulgarian maps which shows the full word блок. The same rule applies for boulevards (shortened to бул.) and streets (usually shortened to ул. or not shown at all).

[lonvia]

Works with new Python frontend.

[Dimitar5555]

@lonvia блок 717 ж.к. Люлин 7 is working correctly on the main OSM website but using the abbreviated version doesn't seem to work (бл. 717 ж.к. Люлин 7
https://www.openstreetmap.org/search?query=%D0%B1%D0%BB.%20717%20%D0%B6.%D0%BA.%20%D0%BB%D1%8E%D0%BB%D0%B8%D0%BD%207).

[lonvia]

Doesn't look like it is a Russian abbreviation known to Nominatim: https://github.com/osm-search/Nominatim/blob/master/settings/icu-rules/variants-ru.yaml

[Dimitar5555]

It's a Bulgarian abbreviation that should be known to Nominatim: https://github.com/osm-search/Nominatim/blob/master/settings/icu-rules/variants-bg.yaml#L4