atx-osg / atx-buildings

Importing Austin buildings and address data into OpenStreetMap
13 stars 2 forks source link

"Half" streets are missing addresses #43

Open talllguy opened 8 years ago

talllguy commented 8 years ago

Greetings ATX OSM team. I was visiting your city this weekend and noticed some peculiarities on the map. There are a number of signed "half" streets, e.g. 38th 1/2 Street or 45th 1/2 Street, in Austin that appear to be introducing some error into your import (or possibly the other way around).

The main issue is that the street name signs (SNS in traffic engineering parlance) usually say E 38th 1/2 St, or something like that. In OpenStreetMap, users have spelled out the 1/2 ordinal to the word "half". I am not sure if this is due to TIGER expansion or some local editor interpretation, but the use of "Half" doesn't reflect the sign.

That aside, it appears your address matching code has ignored addresses along "half" streets throughout Austin. In the example below, you can see the 45th 1/2 Street does not have any addresses.

2016-01-18 14_53_11-openstreetmap http://www.openstreetmap.org/#map=19/30.30620/-97.72033

38th 1/2 Street (recently edited by mapbox) also has no addresses along it

Note: The favored OSM naming for half addresses and streets is to write out the "1/2", and not to use the Unicode character (see https://github.com/osmlab/nycbuildings/issues/67).

wilsaj commented 8 years ago

I think the issue is, as you pointed out, that the OSM street names are not really correct. Most locals would consider that street name to be "East 45th 1/2 Street", and the source dataset from the City of Austin also represents them as such.

In doing the address matching, we filtered out address points where we couldn't determine a good street name match with a high degree of confidence, and in these cases the difference in representation was enough to cause a miss.

Our plan is to finish the validation and cleanup of the easily matched addresses (currently in progress), then go back and tackle the weird cases like these as part of a rolling update strategy. The source address dataset is updated monthly, so a general process for identifying gaps or mis-matches between City of Austin addresses and OSM addresses should be able to help nail these down as well.