RecogniContact Contact Data & Address Parser

Highlights

Main features

Online demo

Products

Countries and languages

For the following countries, RecogniContact recognizes contact information including postal addresses. For these countries, RecogniContact includes a comprehensive database with place names, so that it can identify the country in a postal address even if the country is not explicitly specified.

Countries RecogniContact splits up contact information including postal address data for the following countries and regions: USA Canada Europe (except greek and cyrillic alphabets) Australia

Languages RecogniContact recognizes all commonly used strings for structuring contact information (e.g., name: address: phone: email: etc.) in the following languages: Catalan Danish Dutch English Finnish French German Icelandic Italian Norwegian Portuguese Spanish Swedish

Example: USA postal address
11 West 53 Street New York, NY 10019 United States		Street address= 11 West 53 Street ZIP/Postal code= 10019 Place name= New York State/Region/Province=NY Country= United States ISO country code= US

Example: UK postal address
53 Bankside London SE1 9TG United Kingdom		Street address= 53 Bankside ZIP/Postal code= SE1 9TG Place name= London Country= United Kingdom ISO country code= GB

Example: German postal address
Potsdamer Str. 50 10785 Berlin Deutschland		Street address= Potsdamer Str. 50 ZIP/Postal code= 10785 Place name= Berlin Country= Deutschland ISO country code= DE

For all countries of the world, RecogniContact splits up contact information without postal addresses.

The following preconditions apply:

• The data is noted in Latin letters (as opposed to, for example, Greek or Cyrillic letters).
  
• Language-dependent elements are specified in one of the 13 languages currently supported.
  

Country detection

• ZIP/postal code format and place name (the integrated database comprises more than 200,000 place names)
• Country codes in telephone numbers
• Country domains in email and Web addresses

This information will be used to standardize phone numbers to a unified format or to add country information that is missing from a postal address.

Example: Detect country from address format and place name
Potsdamer Str. 50, 10785 Berlin		Street address= Potsdamer Str. 50 ZIP/Postal code= 10785 Place name= Berlin Country= Germany ISO country code= DE

Example: Detect country from phone number
Phone 004315556666		Phone= +43 (1) 5556666 Country= Austria ISO country code= AT

Format-independent parsing

• if contact data comes from sources (email messages, websites) where the items don't have any predefined structure, or
• if addresses are copied from tabular sources such as spreadsheets or tables on websites.

Example: Unformatted address, no separators
11 West 53 St New York New York 10019		Street address= 11 West 53 St ZIP/Postal code= 10019 Place name= New York State/Region/Province=New York Country= United States ISO country code= US

Contact information fields

RecogniContact extracts the following fields from text containing contact information:

• Person-related fields
• • Name prefixes (Mr., Dr., etc.)
  • First name or initial letter
  • Second name or initial letter
  • Last name
  • Suffix – titles such as Ph.D., MBA or name suffixes such as Junior, Jr.
  • Position (technical director, marketing manager, etc.)
  
  
• Company/organization-related fields
• • Company/organization name
  • Department
  
  
• Address
• • Street address
  • ZIP/postal code of the street address
  • Post office box address
  • ZIP/postal code of the post office box address
  • Place name
  • Country
  • Region information: state (USA), county (Ireland), province (Italy), canton (Switzerland), Bundesland (Germany) …
  
  
• Telephone numbers
• • Fixed line
  • Mobile phone
  • Fax number
  
  
• Internet
• • Email address
  • Website address
  
  

Example: Name parsing
Mr Walter W. Wagoner, PhD		Prefix= Mr First name= Walter Middle name= W. Last name= Wagoner Suffix= PhD Gender= M

Example: Name, position and organization
Walter Wagoner, Exhibition Manager Museum of Modern Art		First name= Walter Last name= Wagoner Company= Museum of Modern Art Position= Exhibition Manager Gender= M

Example: Separate street and post office box addresses
Rozenstraat 59, 1016 NN Amsterdam Postbus 75082, 1070 AB Amsterdam		Company= Street address Street address= Rozenstraat 59 Postbox address= Postbus 75082 Postal code of postbox=1070 AB ZIP/Postal code= 1016 NN Place name= Amsterdam Country= Netherlands ISO country code= NL

Phone number parser

Example: Standardize phone numbers – country prefix, regional prefix
Germany, Tel: 0305556666		Country= Germany Phone= +49 (30) 5556666 ISO country code= DE

Example: Mobile phone numbers
Phone number: +491515556666		Country= Germany Mobile phone= +49 (151) 5556666 ISO country code= DE

Integrated databases of known items

RecogniContact maintains a comprehensive database that contains, among other things, the following information:

• More than 200,000 place-names in the USA, Canada, Australia, and the following European countries:
  Austria, Belgium, Denmark, Finland, France (plus Monaco), Germany, Iceland, Ireland, Italy (plus San Marino and the Vatican), Luxembourg, The Netherlands, Norway, Portugal, Spain (plus Andorra), Sweden, Switzerland (plus Liechtenstein), United Kingdom
  
  This place name database permits the identification of the country according to the postal address given, even if the country is not specified explicitly.


• 
  
• 12,000 first names with gender information.
• 
  
• The international country prefixes of all countries in the world, from +1 (USA & Canada) to +997 (Bahamas).
• 
  
• Multilingual place names such as Vienna, Vienne, Wien, Wenen …
• 
  
• Country-specific regional information, including state (USA), province (Italy), canton (Switzerland), Bundesland (Germany), county (Ireland) …
• 
  
• Strings in 13 languages for the following elements:
  
  
  • Country names
    Germany, Deutschland, Allemagne, Duitsland
  • 
    
  • Job titles
    Director, Direktor, Directeur
  • 
    
  • Common street identifiers
    -street, -straße, rue, -straat
  • 
    
  • Post office box identifiers
    P.O. Box, Postfach, Boîte postale, Postbus
  • 
    
  • Salutations
    Mrs., Fr., Mme, Mevr
  • 
    
  • Company types
    Ltd, GmbH, Sarl, BV
  • 
    
  • Strings used to structure contact information:
    Name, Nom, Naam, Namn 

Person's gender

First names that don't allow a conclusion on the person's gender are taken into consideration: Alex, Cameron, Chris, Sasha.

Example: Gender detection
Walter Wagoner		First name= Walter Last name= Wagoner Gender= M

Structuring elements, unrecognized information

Example: Structuring elements in Finnish
Postiosoite: Mannerheiminaukio 2, 00100 Helsinki Puhelin: +358 9 1733 6501		Street address= Mannerheiminaukio 2 ZIP/Postal code= 00100 Place name= Helsinki Country= Finland Phone= +358 (9) 1733 6501 ISO country code= FI

Example: Unrecognized information
Organisation: Museum of Modern Art Phone: (212) 708-9400 Opening times: 10:30 a.m.–5:30 p.m.		Company= Museum of Modern Art Phone= (212) 708-9400 Unrecognized values: Opening times: 10:30 a.m.–5:30 p.m.

Unmatched recognition accuracy

Recognition algorithms that try to reduce the problem of contact information parsing to a few standard recognition patterns show their limitations very quickly. They are unable to bear the challenges of robust and reliable contact and address parsing in some very common situations:

• When an unexpected address supplement is used
• When unexpected punctuation characters are used in the source text or white spaces are misplaced
• When items are not separated by new line characters or by consistent separator characters
• When elements are copied from spreadsheets in which the data is arranged in a tabular format

Input data as diverse and complex as contact and address information cannot be captured in a limited number of recognition patterns or regular expressions. This is particularly true with international address data. In practice, only a small fraction of all textual contact information complies with standards. It is impossible to create a comprehensive list of all address formats and conventions ever used in reality. And even if such a list were available, at a certain level of complexity merely pattern-based recognition would be very inefficient.

To meet these challenges, LoquiSoft has created a set of algorithms tailored specifically to the problem of address and contact information parsing.

Quality measures

As in all software that processes semantic and language-dependent information, a certain residual error rate cannot be avoided in address parsing, due to unknown or ambiguous information (example: automatic spell-checking).

To ensure the quality of RecogniContact's recognition results, LoquiSoft uses the following methods:

• Test database
  During every software update, a test database with thousands of manually tagged contact data records is used as a basis to verify and improve RecogniContact's parsing algorithms.
• 
  
• Flexible recognition rules
  The set of recognition rules within RecogniContact is highly flexible. New rules, exceptions and exceptions of exceptions can be added without the risk of making the recognition algorithm slow, inefficient or overly complex to manage.
• 
  
• User feedback
  Whenever users of our ContactCopy product find recognition problems, they can report them to us with just a mouse click. This valuable feedback has helped us eliminate problems and improve our recognition algorithms since ContactCopy's original release in 2007.

Reference clients