[Taginfo-dev] Tag synonyms

[Stefan Keller]

Hi Bryce

I strongly agree that there is a need that users can search for tags
via a specialized API (in topical or vertical search engine API).
Unfortunately I know no answer yet within OSM which does that.

Taginfo is definitely a good starting point. And yes, it's a challenge
to extract synonyms (or related terms). But just to complete what
Jochen wrote:

There is already a proposal of putting synonyms (and more) into the
Wiki text! See "Related Terms" at bottom e.g. of "forest" [1].
Everyone can contribute by adding such related terms right now.

Based on Tagfinder db and API we implemented a prototype called
TagFinder which gives you an idea how nice such a function would be:
If you enter "Wald" into TagFinder [2] you get "landuse=forest" in a
ranked result page. This prototype reads out the synonyms from the
Wiki pages mentioned above and includes a thesaurus. For more
background of TagFinder see slide 8 at [3].

Yours, Stefan

[1] http://wiki.openstreetmap.org/wiki/Tag:landuse%3Dforest
[2] http://openpoimap.ch/tagfinder/search?q=Wald
[3] http://www.fossgis.de/konferenz/2012/programm/attachments/328_FOSSGIS%202012%20-%20OpenPOIMap%20-%20en.pdf



2013/1/26 Bryce Cogswell <bryceco at yahoo.com>:
> On Jan 26, 2013, at 12:12 PM, Jochen Topf <jochen at remote.org> wrote:
>
>> On Sat, Jan 26, 2013 at 10:52:06AM -0800, Bryce Cogswell wrote:
>>> I have an editor that consumes TagInfo data from the wiki, allowing users to search for the tag they want. I search both the tag name and the description text. I'd like the search to include common synonyms for tags because newbies (and I) don't always know what to search for, so: drugstore -> amenity=pharmacy, salon -> shop=beauty, curb -> kerb, etc.
>>>
>>> What is the correct place to put this information? Should it be a new tag metadata field on the wiki, or stuffed into the description field, or something else?
>>
>> This mailing list is about the taginfo application/service. From your question
>> it sounds that you are not actually doing anything with taginfo, but parsing
>> the Wiki directly. Is that correct? In that case you might want to raise this
>> issue on the general dev or talk lists. That being said, I do think that the
>> taginfo program is the right place to collect this information and bring it
>> to other parts of the OSM ecosystem.
>>
>> I think framing the issue as being about synonyms is too simplistic. It is
>> about finding the right tags. Just having a list of synonyms is not going to be
>> enough. Say for instance "forest" and "woods". Those are synonyms by some
>> definitions.  But there are actually different tags for these two things,
>> because in the OSM context they mean different things. So you'd have to read
>> the description on the wiki to find out which to use. I am sure there are many
>> words you can't easily associate with just one tag.
>>
>> The other problem is maintainance. People actually have to maintain the lists
>> of synonyms. And there is the internationalization issue of course.
>>
>> So I think the approach has to be different. I suggest a mixture of full-text
>> search in wiki pages plus thesaurus plus maybe some kind of word-stemming plus
>> maybe dictionaries. It is a bit of work to get that going, but I think it
>> can be done. We'll get a lot of synonyms for free that way and where we don't
>> get them, somebody just has to make sure they appear in the wiki text somewhere.
>>
>> There is already a very simple full-text search for the wiki integrated into
>> taginfo that I whipped up in a few hours. Just use the search field and then
>> choose the "Full text" tab. Your first two examples will find the right tags:
>> http://taginfo.openstreetmap.org/search?q=drugstore#fulltext
>> http://taginfo.openstreetmap.org/search?q=salon#fulltext
>> That could be the basis of something better.
>>
>> Jochen
>> --
>> Jochen Topf  jochen at remote.org  http://www.remote.org/jochen/  +49-721-388298
>
> I currently pull the info from taginfo-wiki-db.bz2 at build time but long term I would prefer to get it via an API (or periodically download the bz2 to the client automatically). My editor is for a mobile device so for performance I don't want to hit the server doing a search on every keystroke.
>
> I expect iD is going to want similar functionality so it would be nice to have it centrally located rather than implemented in each client.
>
> I see that Nominatim has similar functionality already:
> http://wiki.openstreetmap.org/wiki/Nominatim/Special_Phrases/EN
> That may be sufficient for me and gives me a place to submit patches (they include shop=drugstore, unfortunately).
>
> Thanks for your feedback and suggestions.
>
> Bryce
>
>
> _______________________________________________
> Taginfo-dev mailing list
> Taginfo-dev at openstreetmap.org
> http://lists.openstreetmap.org/listinfo/taginfo-dev