Is there a script to standardise OSM data?

In general, there is no current tool available to do what you want. What you are looking for is a script that, for everything in the world that can be represented in OpenStreetMap in two or more different ways, pick one of the representations and convert all the others.

If you are just interested in sorting out tagging, then I would recommend the TagTransform plugin for Osmosis. I use this plugin daily in my rendering toolchain, for example to take the multiple different was of tagging a Zebra Crossing and consolidate them into one tag. This makes the stylesheets easier to deal with.

More advanced geometry manipulation is done in a variety of tools, most noticeably during import by osm2pgsql / nomintim, but these are specific-purposed transformations into a specific data format rather than general-purpose routines that output openstreetmap data.

There is nothing stopping such a utility being created, it simply hasn't been done yet. The ultimate tool would let you pick between the representations depending on your needs - for example some people want is_in generated from polygons, but others prefer polygons generated from is_in tags.

You could probably do this sort of thing with PostGIS. You can use Osmosis or osm2pgsql to read from an .osm file, and output to a PostGreSQL database.

Then you can use PostGIS queries to find if an object is within a boundary polygon, or near a place, and add tags as appropriate. You can also collapse areas into points (e.g. make a POI out of a building or so).

Such queries could be scripted and automated if you want. Though I don't know of any ready made scripts to do the processing/standardisation that you describe. As you say, Nominatim does some of this processing, and its source code is available, which might be helpful.

I'm afraid I think the reason no one's answered this question is that, like me, they don't understand it.

I don't know what a script where all data is kept means. I thought you wanted address data, but now I'm just too confused, and I'm afraid others here are as well.

I'm going to take a stab at answering what I think your question is:

The first part seems to be "How can I programatically access OSM data?"

If so, the answer is several fold.

OSM is a database of geographic information and we provide several mechanisms of storing that data. The most popular consumable format of OSM data is the OSM XML file format. Usually when you see a .osm file, that file is in the OSM XML format, which is described in detail at http://wiki.openstreetmap.org/wiki/.osm

The second part seems to be about programatically modifying OSM features to (using your terminology) "standardise them".

The short answer is that we don't provide a mechanism to do this, and you shouldn't consider modifying OSM data in an automated way without having a deep understanding of the data you're modifying (and often not even then).

Let's take your example of modifying addresses, using the is_in to add street or town data.

Speaking just about the US, I know this wouldn't work. There are places where the city/town you might think corresponds to an address doesn't. There's a long and complex story around this having to do with the names used for towns historically being the name of the post office and not the city, but the bottom line is you don't know.

Similarly, you might assume that a building in proximity to a street is addressed to that street, but you can't know that.

And if you make an assumption like that, modifying the data, then you make it more difficult to find the error and correct in the future.

Of course if you know the information for an area, you should modify it yourself, but this is not the same as carrying out an automated edit (ie running a bot). Bots are generally frowned upon because of their negative history with the project. More information on bots can be found here: http://wiki.openstreetmap.org/wiki/Bot