Reducing duplication on the disambiguation page

[zarino]

Today someone in the FMS user support inbox pointed out a particularly good example of a useless disambiguation page on the Bucks CC cobrand (which uses the OSM geocoder) – but similar issues can happen with the Bing geocoder used by fixmystreet.com.

https://www.fixmystreet.buckscc.gov.uk/around?js=1&pc=newton+longville

These two places are almost identical (51.974429,-0.765671 vs 51.97554,-0.770166) – it makes me wonder whether we could auto-disambiguate when the options are within a mile or so of each other?

Meanwhile, searching for a street name returns 9 results, with very little to distinguish them:

https://www.fixmystreet.buckscc.gov.uk/around?js=1&pc=whaddon+road

From what I can tell, there are actually four "Whaddon Road"s near Milton Keynes – but it’s hard to tell that from the disambiguation page.

This is a tricky problem to solve, but I’m wondering whether some sort of filter ("is this result within X miles of any other result? If so, ignore it") might be a good first pass at an improvement? Or whether maybe the disambiguation page could show the options on its own map?

Another first step might be to gather analytics on what proportion on searches result in a disambguation page, and how the presence of a disambiguation page (or the number of results on it?) affects reporting rates?

[dracos]

I raised a couple of these issues with Nominatim (e.g. returning many parts of the same road as they're different Ways), sadly no response. We do explicitly ask for more than just a road name, given the limitation of geocoders, but I realise people will still do it (less so on .com, I imagine). For cobrands, we try and gain access to council list of road names instead, normally unsuccessfully (Zurich has one).

"is this result within X miles of any other result? If so, ignore it" would not work when you have two identical road names in different places, which of course happens. Showing on a map would presumably help, has its own issues.

[davea]

Another user support message about this:

Its feature which invariably follows the listing of an address, which comes up with irrelevant other postcode addresses on the same road, is frustrating, and seems pointless. A simple first part of a postal code (E8 or N1) should be acceptable on its own, when linked to a road. Someone posting a notice cannot be expected to know the full postal code of any address, and some of those which do appeal in this feature are incomplete.