[postgis-users] Various ways to handle addresses in postgresql
Stephen Woodbridge
stephenwoodbridge37 at gmail.com
Sun Jan 10 09:25:13 PST 2021
It all depends on what you want to do with the various tokens. Is "Nest
2 my nesting place" important from the point of view of geocoding? or
not? You can define a grammar to recognize "Building Names" preceding an
address and separate those out from the rest of the address.
@building @house @street @city @prov @country @postal
@house @street @city @prov @country @postal
@house @street @city @prov @postal
Here is an example of setting up 3 variations of an address definition
and the tokens will get compared to all of them with the best match
sorting to the top of the list.
"@name" is a meta definition that can point to an explicit rules set or
another meta definition
If @build is not important then you can assign the related tokens to
standard field that you ignore in the queries. If it is important then
you assign it to an appropriate field in the standardized address table
for this query.
Here is the sample grammar for great britian (ignore the fact it say
Germany, copy and past error)
https://github.com/woodbri/address-standardizer/blob/develop/data/sample/greatbritain.gmr
And here is the sample lexicon used to classify and standardize tokens
for great britian.
https://github.com/woodbri/address-standardizer/blob/develop/data/sample/greatbritain.lex
These can be edited as required.
Be aware that by default most tokens are classified as WORD, which is
ok, but the more specific classification the better it does to
accurately assign tokens to the correct grammar terms.
Also, this thread is a little off topic for postgis, so unless others
are interested in following this we should not continue on the list. So
if you want to use it, then start with the list of steps I posted
previously and read my docs. Geocoding is not easy as I think you have
already seen. I'm willing to help you but you need to get this built on
your system so we can talk about concrete issues and steps you are
having in implementing it rather than potential problems around
geocoding which are many, and many(most?) of them have been dealt in the
code.
Also be aware that geocoding will never be 100% because it is language
processing problem, but I have been able to get 95+% of a reference data
set to be recognized and matched correctly with multiple different data
sets.
-Steve
On 1/10/2021 9:57 AM, Shaozhong SHI wrote:
> Hi, Steve,
>
> Another solution appeals me most is as follows:
>
> Given a space delimited full address line, we can parse it to correct
> BS7666 format.
>
> Something like house number, street, area, city, postcode
>
> E.g., Nest 2 my nesting place 1B Great Avenue Forest Park London WS22 5TT
> Can you enlighten me about that?
>
> Regards
> ,
> David
>
> On Saturday, 9 January 2021, Stephen Woodbridge
> <stephenwoodbridge37 at gmail.com <mailto:stephenwoodbridge37 at gmail.com>>
> wrote:
>
> David,
>
> This is the link to the address standardizer:
> https://github.com/woodbri/address-standardizer
> <https://github.com/woodbri/address-standardizer>
>
> This is a link to all my code that I developed consulting. It
> includes a few SQL geocoders based on the code above. And has some
> README files discussing how to build a geocoder which is the basis
> for how the geocoders work.
>
> https://github.com/woodbri/imaptools.com
> <https://github.com/woodbri/imaptools.com>
>
> this is the geocoder for Tiger data, but the code is essentially
> the same for every country because the when you load country
> specific data into the database it goes into its own table and
> then you standardize that data into stdstreets table and all
> queries are done against the stdstreets table and you only have to
> tweak the address range interpolation function which needs to
> access the source streets table for the geometry and house number
> ranges.
>
> https://github.com/woodbri/imaptools.com/blob/master/sql-scripts/geocoder/prep-tiger-geo-new.sql
> <https://github.com/woodbri/imaptools.com/blob/master/sql-scripts/geocoder/prep-tiger-geo-new.sql>
>
> I would approach this by:
>
> 1. get the address standardizer compiled and installed. I can help
> if you run into problems or have questions.
> 2. load your UK street data into rawdata schema, ideally it would
> be best if we can create a table/view that presents this data as a
> single table where each record represents one side of the street
> and one jurisdiction this may mean that a single record in your
> source data will generate multiple records in this table/view
> (this greatly simplifies the coding and performance later)
> 3. look at the prep-tiger-geo-new.sql file
> 4. create a stdstreets table and standardize your table/view data
> into it
> 5. look at standardization failures and adjust lexicon and grammar
> as needed
> 6. loop back to 4 until good enough
> 7. load functions from prep-tiger-geo-new.sql file and adjust any
> for your data
> 8. try it out!
>
> -Steve
>
>
> On 1/9/2021 10:22 AM, Shaozhong SHI wrote:
>
> Hi, Stephen,
>
> Many thanks. We are interested in it is working with the UK
> addresses.
>
> Please send me the link to this.
>
> Regards,
>
> David
>
> On Sat, 9 Jan 2021 at 15:00, Stephen Woodbridge
> <stephenwoodbridge37 at gmail.com
> <mailto:stephenwoodbridge37 at gmail.com>
> <mailto:stephenwoodbridge37 at gmail.com
> <mailto:stephenwoodbridge37 at gmail.com>>> wrote:
>
> David,
>
> Yup and this is just one a dozens of cases that you have
> to deal
> with. You are dealing with a natural language processing
> problem.
> And you have to deal with human input that has typos and
> abbreviations.
>
> These issues are what the address standardizer fixes. It
> tokenized
> the address and uses the gazette to standardize the terms
> and then
> classifies each term and assigns it to part of the address
> based
> on a grammar.
>
> So there is a simple solution, use my address
> standardizer, it is
> free, MIT license, it has a sample lexicon/ gazette and
> grammar
> for the UK, it is easy to modify these to fit your needs,
> and it
> just works. Oh if you want to do another county it also
> has sample
> files for 25 countries.
>
> Sent from my iPhone
>
> On Jan 9, 2021, at 4:42 AM, Darafei Komяpa Praliaskouski
> <me at komzpa.net <mailto:me at komzpa.net>
> <mailto:me at komzpa.net <mailto:me at komzpa.net>>> wrote:
>
>
> Hello,
>
> People make neural networks for this kind of task:
>
> https://github.com/openvenues/libpostal
> <https://github.com/openvenues/libpostal>
> <https://github.com/openvenues/libpostal
> <https://github.com/openvenues/libpostal>>
>
> сб, 9 сту 2021, 12:40 карыстальнік Shaozhong SHI
> <shishaozhong at gmail.com
> <mailto:shishaozhong at gmail.com>
> <mailto:shishaozhong at gmail.com
> <mailto:shishaozhong at gmail.com>>> напісаў:
>
> Hi, Steve W,
>
> it is easy to parse addresses as tokens. But it is
> difficult
> to put tokens in right columns, due to that the
> same address
> could be expressed with partial address or full
> address.
>
> The same address can be written like, Flat 1 122
> Great Avenue
> London UK, or Flat 1 122 Greet Avenue Central
> London London
> United Kingdom.
>
> When this happens, each address has different
> number of
> tokens, so different numbers of tokens. Is there
> a way to
> deal with this issue so that each token can get
> into right
> column?
>
> Please enlighten me.
>
> Regards,
>
> David
>
> On Sat, 25 Apr 2020 at 05:09, Stephen Woodbridge
> <stephenwoodbridge37 at gmail.com
> <mailto:stephenwoodbridge37 at gmail.com>
> <mailto:stephenwoodbridge37 at gmail.com
> <mailto:stephenwoodbridge37 at gmail.com>>> wrote:
>
> And I have create an address-standardizer
> project here
> https://github.com/woodbri/address-standardizer
> <https://github.com/woodbri/address-standardizer>
>
> <https://github.com/woodbri/address-standardizer
> <https://github.com/woodbri/address-standardizer>> which
> is user
> configurable. I might be over kill is you just
> want to
> strip off the
> number, in which case you might just use a SQL
> regexp
> replace to remove it.
>
> -Steve W
>
> On 4/25/2020 12:04 AM, Stephen Woodbridge wrote:
> > PostGIS has address_standardizer extension
> that includes
> > parse_address() and standardize_address()
> functions.
> >
> > -Steve W
> >
> > On 4/24/2020 9:54 PM, Imre Samu wrote:
> >> > handle addresses in postgresql
> >>
> >> maybe you can use the
> https://github.com/openvenues/libpostal
> <https://github.com/openvenues/libpostal>
> <https://github.com/openvenues/libpostal
> <https://github.com/openvenues/libpostal>> library
> >> with your favorite language bindings (
> Python / Ruby /
> Go / PHP /
> >> Node / R / Java ...)
> >>
> >> or as a Postgres database extension:
> >>
> https://info.crunchydata.com/blog/quick-and-dirty-address-matching-with-libpostal
> <https://info.crunchydata.com/blog/quick-and-dirty-address-matching-with-libpostal>
>
> <https://info.crunchydata.com/blog/quick-and-dirty-address-matching-with-libpostal
> <https://info.crunchydata.com/blog/quick-and-dirty-address-matching-with-libpostal>>
>
> >>
> >> https://github.com/pramsey/pgsql-postal
> <https://github.com/pramsey/pgsql-postal>
> <https://github.com/pramsey/pgsql-postal
> <https://github.com/pramsey/pgsql-postal>>
> >>
> >> Regards,
> >> Imre
> >>
> >>
> >>
> >>
> >> Shaozhong SHI <shishaozhong at gmail.com
> <mailto:shishaozhong at gmail.com>
> <mailto:shishaozhong at gmail.com
> <mailto:shishaozhong at gmail.com>>
> >> <mailto:shishaozhong at gmail.com
> <mailto:shishaozhong at gmail.com>
> <mailto:shishaozhong at gmail.com
> <mailto:shishaozhong at gmail.com>>>> ezt írta (időpont:
> 2020. ápr. 25.,
> >> Szo, 2:49):
> >>
> >> I find this is a simple, but important
> question.
> >>
> >> How best to split numbers and the rest
> of address?
> >>
> >> For instance, one tricky one is as follows:
> >>
> >> 21-1 Great Avenue, a city, a country,
> this planet
> >>
> >> How to turn this into the following:
> >>
> >> column 1, column 2
> >>
> >> 21-1 Great Avenue, a city, a
> country, this planet
> >>
> >> Note: there is a hyphen in 21-1
> >>
> >> Any clue?
> >>
> >> Regards,
> >>
> >> Shao
> >> _______________________________________________
> >> postgis-users mailing list
> >> postgis-users at lists.osgeo.org
> <mailto:postgis-users at lists.osgeo.org>
> <mailto:postgis-users at lists.osgeo.org
> <mailto:postgis-users at lists.osgeo.org>>
> <mailto:postgis-users at lists.osgeo.org
> <mailto:postgis-users at lists.osgeo.org>
> <mailto:postgis-users at lists.osgeo.org
> <mailto:postgis-users at lists.osgeo.org>>>
> >>
> https://lists.osgeo.org/mailman/listinfo/postgis-users
> <https://lists.osgeo.org/mailman/listinfo/postgis-users>
>
> <https://lists.osgeo.org/mailman/listinfo/postgis-users
> <https://lists.osgeo.org/mailman/listinfo/postgis-users>>
> >>
> >>
> >> _______________________________________________
> >> postgis-users mailing list
> >> postgis-users at lists.osgeo.org
> <mailto:postgis-users at lists.osgeo.org>
> <mailto:postgis-users at lists.osgeo.org
> <mailto:postgis-users at lists.osgeo.org>>
> >>
> https://lists.osgeo.org/mailman/listinfo/postgis-users
> <https://lists.osgeo.org/mailman/listinfo/postgis-users>
>
> <https://lists.osgeo.org/mailman/listinfo/postgis-users
> <https://lists.osgeo.org/mailman/listinfo/postgis-users>>
> >
>
> _______________________________________________
> postgis-users mailing list
> postgis-users at lists.osgeo.org
> <mailto:postgis-users at lists.osgeo.org>
> <mailto:postgis-users at lists.osgeo.org
> <mailto:postgis-users at lists.osgeo.org>>
> https://lists.osgeo.org/mailman/listinfo/postgis-users
> <https://lists.osgeo.org/mailman/listinfo/postgis-users>
>
> <https://lists.osgeo.org/mailman/listinfo/postgis-users
> <https://lists.osgeo.org/mailman/listinfo/postgis-users>>
>
> _______________________________________________
> postgis-users mailing list
> postgis-users at lists.osgeo.org
> <mailto:postgis-users at lists.osgeo.org>
> <mailto:postgis-users at lists.osgeo.org
> <mailto:postgis-users at lists.osgeo.org>>
> https://lists.osgeo.org/mailman/listinfo/postgis-users
> <https://lists.osgeo.org/mailman/listinfo/postgis-users>
>
> <https://lists.osgeo.org/mailman/listinfo/postgis-users
> <https://lists.osgeo.org/mailman/listinfo/postgis-users>>
>
> _______________________________________________
> postgis-users mailing list
> postgis-users at lists.osgeo.org
> <mailto:postgis-users at lists.osgeo.org>
> <mailto:postgis-users at lists.osgeo.org
> <mailto:postgis-users at lists.osgeo.org>>
> https://lists.osgeo.org/mailman/listinfo/postgis-users
> <https://lists.osgeo.org/mailman/listinfo/postgis-users>
>
> <https://lists.osgeo.org/mailman/listinfo/postgis-users
> <https://lists.osgeo.org/mailman/listinfo/postgis-users>>
>
> _______________________________________________
> postgis-users mailing list
> postgis-users at lists.osgeo.org
> <mailto:postgis-users at lists.osgeo.org>
> <mailto:postgis-users at lists.osgeo.org
> <mailto:postgis-users at lists.osgeo.org>>
> https://lists.osgeo.org/mailman/listinfo/postgis-users
> <https://lists.osgeo.org/mailman/listinfo/postgis-users>
> <https://lists.osgeo.org/mailman/listinfo/postgis-users
> <https://lists.osgeo.org/mailman/listinfo/postgis-users>>
>
>
> _______________________________________________
> postgis-users mailing list
> postgis-users at lists.osgeo.org
> <mailto:postgis-users at lists.osgeo.org>
> https://lists.osgeo.org/mailman/listinfo/postgis-users
> <https://lists.osgeo.org/mailman/listinfo/postgis-users>
>
>
> _______________________________________________
> postgis-users mailing list
> postgis-users at lists.osgeo.org <mailto:postgis-users at lists.osgeo.org>
> https://lists.osgeo.org/mailman/listinfo/postgis-users
> <https://lists.osgeo.org/mailman/listinfo/postgis-users>
>
>
> _______________________________________________
> postgis-users mailing list
> postgis-users at lists.osgeo.org
> https://lists.osgeo.org/mailman/listinfo/postgis-users
More information about the postgis-users
mailing list