It's a long time ago, but I'll try to remember how this address de-duplication program operated. The addresses were first parsed into an "internal (canonical) form": for example, "Av" and "Avenue" above would parse to the same internal number. The "internal forms" were compared, not the original text. Addresses that were too weird and couldn't be parsed were rejected for inspection by a human. As part of the parsing, suburbs were checked against an Australian postcode database and errors reported, so the postcodes were cleaned (sometimes by a human). Suburbs were then ignored, postcodes were considered more reliable (people sometimes lie about their suburb name in Australia because some suburbs are "posher" than others).

There was also a user-configurable "number of transcription errors" allowed; if you allowed one transcription error, then "Broadway" and "Brodway" above would be considered a match. There was another parameter for transposition errors (e.g. "Gergory" versus "Gregory") and many other heuristics, like "Flat 6" versus "Flat 6A" versus "Apartment 6". Update: Perhaps a better technique, which I didn't know about at the time and as suggested by metaperl, is Jaro-Winkler distance. I remember trying classical Knuth Soundex and found it to be not useful in this domain.

One of the many imperfect and sometimes amusing heuristics I remember was "Captain Cook Cruises": is this a gentleman with a title of Captain, first name Cook, last name Cruises, or is it a company? Sydneysiders will know the likely answer :-). How about "Mr William and Mrs Melinda Gates": is this "Mr William" and "Mrs Melinda Gates" or "Mr William Gates" and "Mrs Melinda Gates"?

Update: No matter what (imperfect) heuristics you use for the address parsing and matching, you still face the infeasibility of comparing each address with every other one, as n becomes large. An obvious way to tackle that combinatorial explosion is to partition the addresses into "buckets" and compare only addresses within the same bucket. While various "bucketing" schemes are possible, the best/simplest I found was to bucket on postcode.

In reply to Re^3: Efficient Fuzzy Matching Of An Address by eyepopslikeamosquito
in thread Efficient Fuzzy Matching Of An Address by Limbic~Region

Title:
Use:  <p> text here (a paragraph) </p>
and:  <code> code here </code>
to format your post, it's "PerlMonks-approved HTML":


  • Posts are HTML formatted. Put <p> </p> tags around your paragraphs. Put <code> </code> tags around your code and data!
  • Titles consisting of a single word are discouraged, and in most cases are disallowed outright.
  • Read Where should I post X? if you're not absolutely sure you're posting in the right place.
  • Please read these before you post! —
  • Posts may use any of the Perl Monks Approved HTML tags:
    a, abbr, b, big, blockquote, br, caption, center, col, colgroup, dd, del, details, div, dl, dt, em, font, h1, h2, h3, h4, h5, h6, hr, i, ins, li, ol, p, pre, readmore, small, span, spoiler, strike, strong, sub, summary, sup, table, tbody, td, tfoot, th, thead, tr, tt, u, ul, wbr
  • You may need to use entities for some characters, as follows. (Exception: Within code tags, you can put the characters literally.)
            For:     Use:
    & &amp;
    < &lt;
    > &gt;
    [ &#91;
    ] &#93;
  • Link using PerlMonks shortcuts! What shortcuts can I use for linking?
  • See Writeup Formatting Tips and other pages linked from there for more info.