Talk:Name finder

From OpenStreetMap Wiki
Jump to: navigation, search

Character mappings

Please consider additional equivalence mappings for Slavic languages: č=c ć=c š=s ž=z

Example: http://www.frankieandshadow.com/osm/?find=Celovška+cesta is found nicely, but not with ommited accent: http://www.frankieandshadow.com/osm/?find=Celovska+cesta

  • this was done after May 16. Virtually all 'latin-like' characters from the UTF-8 character set are now mapped onto their non-accented equivalents. In some cases alternate forms are also searched: ü is found with u or ue. David.earl 10:41, 24 May 2007 (BST)

Alternative "road" and "street" mappings

"cesta" (Slovenian)

"ulica" (Slovenian)

  • this is a big can of worms, because Slovenian is not the only alternative. Ideally the input and output should both be multilingual. On the other hand, there's probably not a lot of point in allowing Station Road to match Bahnhofstrasse. What we need for the above is (in French for example) 'Rue' and its contraction 'R' to match, but we don't need that to be equivalent to 'road' or 'street'

Ultimatively "Celovška cesta" should be found by http://www.frankieandshadow.com/osm/?find=celovska

  • it now is: see above David.earl 10:46, 24 May 2007 (BST)

Unicode adds a further twist to this, there are often two and sometimes more ways to express the same linguistic unit as Unicode code points. The most relevant example is 'combining forms'. Instead of using U+0161 the Latin small s with caron, you can use U+0073 (an ordinary Latin small s) and U+030C (a combining caron). I suggest that you add a pre-processing step for user queries which converts the incoming Unicode to "Normalization Form C", http://unicode.org/reports/tr15/ before any other step although you might decide that NFKC is a good choice instead. You might also want to pre-process the OSM dump in this way. In the longer term the OSM project should probably agree on a single Normalization Form for the whole project, and again I would suggest NFC.

Example: http://www.frankieandshadow.com/osm/?find=Celovška does not work whereas http://www.frankieandshadow.com/osm/?find=Celovška does work, from a user's point of view they are the same thing, they look the same on screen if you have the right fonts installed for example. Tialaramex 16:49, 28 May 2007 (BST)

Hah, defeated by the Wiki, it converts my examples to NFC so that they are no longer different. Hopefully the explanation is useful at least Tialaramex 20:19, 30 May 2007 (BST)

Source Code availibility

Do you plan to publish source code? User:Zdeno

  • once it settlesdown a bit David.earl 10:39, 24 May 2007 (BST)

Search engine bait

Someone should write a utility to export the planet as HTML files (say 1 per suburb), e.g.

< h1>London< /h1>

< b>Some suburb< /b> near another suburb and another suburb< br>

<a href="http://www.openstreetmap.org/index.html?lat=...&lat=...">Downing Street</a>< br>

...

Then we can host the files somewhere and the search engines can pick it up and allow people to use them as name finder. This will be especially useful for cities without (or with poor) Google Maps coverage but good OSM coverage.

It will also help mapper to debug their work, e.g. finding spelling mistakes -- Nic 22:05, 14 May 2007 (BST)

Well there's this old page: User:LA2/Gazetteer
And no doubt User:Ojw had search engines in mind when he set up http://almien.co.uk/City/
-- Harry Wood 13:16, 31 October 2007 (UTC)

Many language versions

Rather than try to make the site language neutral, which is never going to be totally possible anyway since the UI and results are English, it ought to be possible to make it localisable, ie so that the same program code can drive a site written in Japanese, or Hebrew, or Swedish by using suitably localised data.

Basically at a first approximation this means factoring out human readable strings like 'which is about %d km from' and 'near' from the code, and replacing them with configurable variables. Then use the same language setting to choose the preferred language of streets, towns etc. found in the OSM database. Since this name finder is so popular it should be easy to find volunteers for one or two non-English languages to translate these strings and see how well it works. Tialaramex 16:49, 28 May 2007 (BST)

Context

Getting context from smaller places doesn't make sense. eg. City, about 20km north of town. I think context should be restricted to places of equal or larger size.

Also, I think the is_in attribute should be preferred to a nearby place, especially if there are no places within a large distance. Recursive is_in attributes is nice too (eg. city, in state, country, continent.) Justcameron 18:12, 22 September 2007 (BST)

Amenities with no name not found

For some reason name finder doesn't find amenities with no name. (inner instead of outer join in SQL?)

eg hotel near ljubljana finds:

  • hotel Slon found less than 1km south of middle of city; traffic signals Ljubljana
  • hotel Lev Intercontinental found less than 1km north-west of middle of city; traffic signals Ljubljana
  • hotel M Hotel found less than 1km south-east of middle of suburb Šiška (which is about 2km north-west of middle of city; traffic signals Ljubljana

but misses several hotels in the city centre, that have no names in the OSM db, all only a block away (200m NE) from hotel Slon (1st match), and less than 1km southeast of the city centre (on Miklošičeva cesta) --Stefanb 09:40, 2 October 2007 (BST)

OK, seems fixed now. -- Stefanb 23:38, 18 November 2007 (UTC)

street numbers

Street numbers in a search cause rather unexpected failures. Given that we don't have street numbers in general, I'd like the numbers to be ignored. This makes it easier for a "show on map" link of an address to do something useful with OSM. Currently searching for "Hverfisgata, Reykjavik, Iceland" shows you the street. searching for "34 Hverfisgata..." or "Hverfisgata 34..." gives no results at all, not even just Reykjavik.

+1 vote from me -- Stefanb 23:36, 18 November 2007 (UTC)

Getting up to date...

Why this index is updated so rarely? Last update is from 17 Oct 2007 - now is 11 Nov, any plans for more rapid updates? Sauron

That is something that annoys me, too. And since you wrote until today has passed more than one year! --Geogast 10:02, 12 March 2009 (UTC)
"Index from OpenStreetMap data up to 9 Jan 2009" (http://gazetteer.openstreetmap.org/namefinder/) Holy Mercator, that's not quite up-to-date. :( --Nix78 11:22, 6 November 2009 (UTC)

reindexing means no results on main page search

When the namefinder is being reindexed, searchs from the main search box on the OSM homepage return no results (unless your search also exists in geonames) There's also no message indicating that search is being reindexed. This was rather offputting, as I knew things used to work. I couldn't work out whether I simply hadn't mapped the street I was trying to find, or what had happened. Kpalsson 22:09, 29 February 2008 (UTC)

How to change useless descriptions?

Often name finder descripes places a little bit strange. An example:

town Hof in Oberfranken, Bayern, Bundesrepublik Deutschland, Europe which is about 11km north of town Schwarzenbach an der Saale in Hof, Oberfranken, Bayern, Bundesrepublik Deutschland, Europe

This description may be usefull for all those, who know the little town Schwarzenbach an der Saale with its 8.000 people. But almost everyone of them would also know the next bigger town Hof with 50.000 people. So this description is useless for almost all users.

How can it be changed? For exampe to:

Hof ... midway between Nuremberg and Leipzig

thanks for comments. --Wst 05:57, 10 May 2008 (UTC)

Guide for deploying an own namefinder?

Hi,

I am in the process of building a local OSM application server, that should one day provide offline access to OSM maps in case of emergency for our fire brigade (our IT-equipment will be powered from generators, then) , so I am interested in running my own namefinde on that box. I've got already PostGIS and Mapnik running with an OpenLayers interface, update scripts that would pull OSM XML data every two weeks and running a rerender of the map for a small region of germany.

So I'd like to run a namefinder service on the box, could someone point out how to set it up after getting the site code from svn? MaySQL5 and PHP5 is obviously required, but how to hook it all up in a simple way including update scripts? I'm on Debian Etch.

Any help would be greatly appreciated!--Thorongil 13:55, 8 October 2008 (UTC)

When are old ways reindexed?

I've changed the name of a part street in Cologne, Germany from "Alter Niehler Kirchweg" to "Niehler Kirchweg" some days ago.

But even after some days, the namefinder shows "Alter Niehler Kirchweg" in the search results.


How often is the index updated, and when will this old way removed?

Just a lat/long seems to fail

I get an internal error when providing just a lat/long pair as shown in the examples.

I can't come up with a lat/long pair that works. I'm going to delete this example until it's working. --Evan Prodromou 20:34, 9 December 2009 (UTC)

OK, changed my mind. Since "places near 52.18,0.20" seems to work, it's probably better to use that than to delete the entry entirely. --Evan Prodromou