NAME

Lingua::EN::AddressParse - extract components of a street address from free format text

SYNOPSIS

    use Lingua::EN::AddressParse;

    my %args =
    (
      country     => 'US',
      auto_clean  => 1,
      force_case  => 1,
      abbreviate_subcountry => 0,
      abbreviated_subcountry_only => 0,
      force_post_code => 0
    );

    my $address = Lingua::EN::AddressParse->new(%args);
    $error = $address->parse("40 1/2 N OLD MASSACHUSETTS AVE APT 3B Washington Valley Washington 98100: HOLD MAIL");
       
    print $address->report;       
       
        Country address format  'US'
        Address type            'suburban'
        Non matching part       'HOLD MAIL '
        Error                   '1'
        Error descriptions      'non matching section : HOLD MAIL '
        Warning                 '1'
        Warning description     ''
        Case all                '40 1/2 N Old Massachusetts Ave Apt 3B Washington Valley WA 98100'
        COMPONENTS              ''
        base_street_name        'Old Massachusetts'
        post_code               '98100'
        property_identifier     '40 1/2'
        street_direction_prefix 'N'
        street_name             'N Old Massachusetts'
        street_type             'Ave'
        sub_property_identifier '3B'
        sub_property_type       'Apt'
        subcountry              'WASHINGTON'
        suburb                  'Washington Valley'       

    %address_components = $address->components;
    print $address_components{sub_property_type};       # APT
    print $address_components{sub_property_identifier}; # 3B
    print $address_components{property_identifier};     # 40 1/2
   
    %address_properties = $address->properties;
    print $address_properties{type};            # suburban
    print $address_properties{non_matching};    # : HOLD MAIL

    $correct_casing = $address->case_all;

DESCRIPTION

This module takes as input a suburban, rural or postal address in free format text such as,

    3080 28TH AVE N ST PETERSBURG, FL 33713-3810
    12 1st Avenue N Suite # 2 Somewhere CA 12345 USA
    C/O JOHN, KENNETH JR POA 744 WIND RIVER DR SYLVANIA, OH 43560-4317

    9 Church Street, Abertillery, Mid Glamorgan NP13 1DA
    27 Bury Street, Abingdon, Oxfordshire OX14 3QT

    2A O'CONNELL ST KEW NSW 2123
    12/3-5 AUBREY ST MOUNT VICTORIA VICTORIA 3133
    "OLD REGRET" WENTWORTH FALLS NSW 2782 AUSTRALIA
    GPO Box K318, HAYMARKET, NSW 2000

and attempts to parse it. If successful, the address is broken down into it's components and useful functions can be performed such as :

    converting upper or lower case values to title case (2A O'Connell St Kew NSW 2123)
    extracting the addresses individual components      (2A,O'Connell,St,KEW,NSW,2123)
    determining the type of format the address is in    ('suburban')

If the address cannot be parsed you have the option of cleaning the address of bad characters, or extracting any portion that was parsed and the portion that failed.

This module can be used for analysing and improving the quality of lists of residential and postal addresses.

By using a large combination of regular expressiosn with look ahead analysis, patterns can be parsed that confuse many other parsers. Examples are

Street names with several street types: Lane Cove Road Suburbs which include street types: Smith Road St Marys Suburbs that include state names: Fort Washington Washington

DEFINITIONS

The following terms are used by AddressParse to define the components that can make up an address.

    Pre cursor : C/O MR A Smith...
    Sub property identifier : Level 1A Unit 2, Apartment B, Lot 12, Suite # 12 ...
    Property Identifier : 12/66A, 24-34, 2A, 23B/12C, 12/42-44, 2.5
    Property name   : "Old Regret"
    Post Box        : GP0 Box K123, LPO 2345, RMS 23 ...
    Road Box        : RMB 24A, RMS 234 ...
    Street Direction: North, SE, Sth. etc
    Street name     : O'Hare, New South Head, The Causeway, Broadway
    Street type     : Road, Rd., St, Lane, Highway, Crescent, Circuit ...
    Suburb          : Dee Why, St. John's Wood ...
    Sub country     : NSW, New South Wales, ACT, NY, New Jersey AZ ...
    Post (zip) code : 2062, 34532-1234, SG12A 9ET
    Country         : Australia, UK, US or Canada

The main address formats currently supported are as follows. (a ? means the component is optional):

    'suburban' : sub_property(?) property_identifier(?) street street_type suburb subcountry post_code(?)country(?)

    OR for the USA
    'suburban' : property_identifier(?) street street_type sub_property(?) suburb subcountry post_code(?) country(?)

    'rural'    : property_name suburb subcountry post_code(?) country(?)
    'post_box' : post_box suburb subcountry post_code(?) country(?)
    'road_box' : road_box street street_type suburb subcountry post_code(?) country(?)
    'road_box' : road_box suburb subcountry post_code(?) country(?)

Note that suburb and subcountry are not optional. The accuracy of the parser is improved by providing as much context as possible. Proding a suburb can ehlp to identify street names that would itherwise be ambigious.

For the case where you only have a street address, dummy (but still valid) values can be used for suburb (such as 'Somewhere') and sub country (such as 'NY'). These dummy values will be parsed but can be ignored.

All formats may contain a precursor

Refer to the component grammar defined in the Lingua::EN::AddressParse::Grammar module for a complete list of combinations.

METHODS

new

The "new" method creates an instance of an address object and sets up the grammar used to parse addresses. This must be called before any of the following methods are invoked. Note that the object only needs to be created once, and can be reused with new input data.

Various setup options may be defined in a hash that is passed as an optional argument to the "new" method.

    my %args =
    (
      country     => 'US',
      auto_clean  => 1,
      force_case  => 1,
      abbreviate_subcountry => 1,
      abbreviated_subcountry_only => 1,
      force_post_code => 1
    );

    my $address = Lingua::EN::AddressParse->new(%args);

country

The country argument must be specified. It determines the possible list of valid sub countries (states, counties etc, defined in the Locale::SubCountry module) and post code formats. Either the full name or abbreviation may be specified. The currently supported country names and codes are:

    AU or Australia
    CA or Canada
    GB or United Kingdom
    US or United States

All forms of upper/lower case are acceptable in the country's spelling. If a country name is supplied that the module doesn't recognise, it will die.

force_case (optional)

This option only applies to the "case_all" method, see below.

auto_clean (optional)

When this option is set to a positive value, the input string is 'cleaned' to try and normalise bad patterns. The type of cleaning includes

    remove non alphanumeric characters
    remove full stops
    remove redundant white space
    add missing space separators
    expand abbreviations to more common forms
    remove bracketed annotations
    fix badly formed sub property identifiers

abbreviate_subcountry (optional)

When this option is set to a positive value, the sub country is forced to it's abbreviated form, so "New South Wales" becomes "NSW". If the sub country is already abbreviated then it's value is not altered.

abbreviated_subcountry_only (optional)

When this option is set to a positive value, only the abbreviated form of sub country is allowed, such as "NSW" and not "New South Wales". This will make parsing quicker and ensure that addresses comply with postal standards that normally permit only abbreviated sub countries.

It also avoids matching a sub_country name too early, as in the case of 'Port Washington New Jersey' Normally, 'Washington would be consumed as the sub country, but by first converting the address to 'Port Washington NJ' we avoid this problem

force_post_code (optional)

When this option is set to a positive value, the address must contain a post code. If it does not then an error flag is raised. If this option is set to 0 than a post code is optional.

By default for this option is true.

parse

    $error = $address->parse("12/3-5 AUBREY ST VERMONT VIC 3133");

The "parse" method takes a single parameter of a text string containing a address. It attempts to parse the address and break it down into the components described below. If the address is parsed successfully, a 0 is returned, otherwise a 1.

Note that you can successfully parse all the components of an address and still have an error returned. This occurs when you have non matching data following a valid address. To check if the data is unusable, you also need to use the "properties" method to check the address type is 'unknown'

This method is a prerequisite for all the following methods.

components

    %address = $address->components($upper_case_all);
    $suburb = $address{suburb};

If the optional argument $upper_case_all is set to a postive value, all components are converted to upper case.

The "components" method returns all the address components in a hash. The following keys are used for each component:

    pre_cursor - such as 'C/O Mr A Smith'
    po_box_type - such as 'Private Boxes'
    post_box
    road_box
    sub_property_type
    sub_property_identifier
    property_identifier
    property_name
    level - such as 12th Floor
    building - such as Tower A
    street_direction_prefix (such as East, NW, North etc)
    base_street_name (the name with direction removed, such as "Main" in "East Main St")
    street_name (the full street name such as "East Main")
    street_type
    street_direction_suffix (US only, abbreviated only such as N, SE etc)
    suburb
    subcountry
    post_code
    country

If a component has no matching data for a given address, it's values will be set to the empty string.

Each component is converted to title case, meaning the first letter of each component is set to capitals and the remainder to lower case.

Proper name capitalisations such as MacNay and O'Brien are observed

The following components are not converted to title case:

    post_box
    road_box
    subcountry
    post_code
    country
    street_direction_suffix

If your input data is all upper case and you want to retian that format for parsed data, you will need to apply the 'uc' function to each component.

case_all

    $correct_casing = $address->case_all;

The "case_all" method does the same thing as the "components" method except the entire address is returned as a title cased text string.

If the force_case option was set in the "new" method above, address case the entire input string, including any unmatched sections after a recognisable address that failed parsing. This option is useful when you know you have invalid data, but you still want to title case what you have.

properties

The "properties" method returns several properties of the address as a hash. The following keys are used for each property -

    type - either suburban ,rural,post_box,road_box,unknown
    non_matching - any trailing string not part the address

Additional properties can be accessed with the following

  $address->{original_input}
  $address->{input_string} - string after auto_clean option has been applied
  $address->{country_code} - abbreviated Country address format (as defined in the C<new> method)
  $address->{error} - error flag, 0 = good, 1 = error
  $address->{error_desc} - text to describe the type of parsing error
  $address->{warning} - warning flag, 0 = good, 1 = warning
  $address->{warning_desc} - text to to describe the type of parsing warning(s)

Warnings mean that the address has parsed but there may still be errors within it's components

report

Create a formatted text report

    the input string
    the cleaned input string
    the country type
    the address type
    any non matching part of input string
    if any parsing errors occurred
    error description
    if any parsing warning occurred
    warning description

    the name and value of each defined component

Returns a string containing a multi line formatted text report

DEPENDENCIES

Lingua::EN::NameParse, Locale::SubCountry, Parse::RecDescent

BUGS

LIMITATIONS

Streets such as 'The Esplanade' will return a street of 'The Esplanade' and a street type of null string.

The abbreviation 'St' can be interpreted as either street or Saint. This leads to ambiguities such as '12 East St Thomas Lane'. This could be 'East Street', suburb of 'Thomas Lane' or 'East St Thomas Lane'. And the first pattern is the more common, that is what will match.

For US addresses, an ambiguity arises between a street directional suffix and a suburb directional prefix, such as '12 Main St S Springfield CA 92345'. Is it South Main St, or South Springfield? The parser assumes that 'S' belongs to the street description.

The huge number of character combinations that can form a valid address makes it is impossible to correctly identify them all.

Valid addresses must contain:

    property address, suburb, subcountry (aka state) in that order.

This format is widely accepted in Australia and the US.

UK addresses will often include suburb, town, city and county, formats that are very difficult to parse.

Property names must be enclosed in single or double quotes like "Old Regret"

Because of the large combination of possible addresses defined in the grammar, the program is not very fast.

REFERENCES

"The Wordsworth Dictionary of Abbreviations & Acronyms" (1997)

Australian Standard AS4212-1994 "Geographic Information Systems - Data Dictionary for transfer of street addressing information"

ISO 3166-2:1998, Codes for the representation of names of countries and their subdivisions. Also released as AS/NZS 2632.2:1999

REPOSITORY

<https://github.com/kimryan/Lingua-EN-AddressParse>

TO DO

Define grammar for other languages. Hopefully, all that would be needed is to specify a new module with its own grammar, and inherit all the existing methods. I don't have the knowledge of the naming conventions for non-english languages.

AUTHOR

AddressParse was written by Kim Ryan <kimryan at cpan d o t org>

COPYRIGHT AND LICENSE

This library is free software; you can redistribute it and/or modify it under the same terms as Perl itself.