There is no point in storing the rating score: it changes and there are
plenty of other places where this can be found (I think the FHRS have an
android app themselves. The primary use is to add address data. A secondary
use is to spot places which are no longer in business.

In general do not add unmapped FHRS nodes unless you have other information
which indicates the place is open (even an inspection 2 months ago may be a
place which has closed). I've added a couple recently based on news reports.

I would suggest that using something like fhrs_ref is best. I sometimes
link the url as a source (source:website=) but this is merely to assist
others in verifying the data.

FHRS data varies a lot in quality between local authorities. Powys seem to
use very non-standard addresses, and Rotherham don't seem to be bothered
with post codes. Windsor & Maidenhead has the most useful descriptions,
particularly for canteens (Compass Catering at Sainsbury's). There is scope
to get your local authority to improve consistency of data collection.

The lead time for new establishments to appear on FHRS seems rather long. A
new takeaway opened in June and still does not appear in the extracts. I
know some businesses who wait for the FHRS inspection before starting to
trade, but don't know what the general picture is. Lots of places dont seem
to be covered, including the place in Kenton mentioned on this list a
couple of months ago (I queried both Harrow & Brent about it).

I hope in the near future to be able to provide:

   - Regular diffs of FHRS data.
   - Full CSV version of current data.
   - Archive copies of the xml.
   - Slippy map tiles with geolocated FHRS data shown based on the postcode
   centroid (so no Rotherham data)

I'm just working to tidy up my shell script which does the first two, and
Chris Fleming has offered to host the script (I don't have a linux
environment, let alone one running 24/7), so that we have a reasonable run
of changes starting from 1st October. I had grandiose ideas of aligning the
FHRS nodes by house number in the right direction along the road, but until
I work through a host of interesting issues, will just have to make do with
a less sophisticated version.

The full UK FHRS XML data zipped is around 80M (350 files, 400Mb, my CSV
extract (13 columns, 440k) is also around 80Mb (16.5Mb zipped). Daily diffs
are about 150kb (~ rows).

I plan to push the data in to Postgres & create a temporalised table for
the data, preferably with change detection on a by column basis. I have
encountered some encoding issues with some data from Northern Ireland (they
seem to 1\2 to mean housenumber 1 to 2 inclusive).

A more challenging project is to work on conflation of FHRS data with OSM
data. Names and addresses are rarely perfect matches with OSM. Geolocation
of the FHRS data is only to postcode centroid. I believe we need to be able
to integrate some likelihood estimate of match based on the following
independent criteria:

   - Type of POI bearing in mind that things like pubs might be mapped as
   restaurants and vice versa.
   - Establishment Name. Individual elements of the name probably should be
   treated as tokens for matching (so we match Sycamore Primary School to
   Sycamore Academy; The Rose & Crown Inn with Rose and Crown).
   - Address. Clearly Nominatim can do much of this, but if the address
   needs to be parsed then complications ensure.
   - Postcode. There are errors in postcodes, usually single letter
   transpositions. Certain postcodes may have the postcode centroid a long way
   from the POI, but in most cases they will be within 100m.
   - It is to be hoped that the Local Authority is unambiguous!

Anyone interested in taking this latter aspect further?

Regards,


Jerry




On 19 October 2013 14:22, Neil Pilgrim <[email protected]> wrote:

> I've used fhrs to add some data and wondered about this, though in
> Scotland they didn't seem to have a rating. I wondered if the fhrs ref was
> unique so we could use eg fhrs_ref=[number]. I wasn't sure which key in the
> xml to use at first glance, but hadn't really looked in detail. I've used
> source:addr=fhrs or fhrs in plain source tag too.
>
> thanks to sk53 for pointing out this source of information, it looks
> really promising :)
>
> Neil
>
> On 19 Oct 2013 14:01, "Dave F." <[email protected]> wrote:
> >
> > Hi
> >
> > I wanted to add url links for UK Food Hygiene Rating System to
> restaurants & cafes etc. Such as:
> >
> > http://ratings.food.gov.uk/business/en-GB/516821
> >
> > There's an OSM wiki page but that's just information about the database.
> > There's also a few tags of fhrs:hygiene but these give values (eg 1-5)
> which can change as hygiene inspections are always ongoing.
> >
> > How's best to go about this? Is there a tag already in use that I can't
> find?
> >
> > Dave F.
> >
> > _______________________________________________
> > Talk-GB mailing list
> > [email protected]
> > https://lists.openstreetmap.org/listinfo/talk-gb
>
> _______________________________________________
> Talk-GB mailing list
> [email protected]
> https://lists.openstreetmap.org/listinfo/talk-gb
>
>
_______________________________________________
Talk-GB mailing list
[email protected]
https://lists.openstreetmap.org/listinfo/talk-gb

Reply via email to