Do you know where your variables are? | |
PerlMonks |
comment on |
( [id://3333]=superdoc: print w/replies, xml ) | Need Help?? |
We geo code all addresses then use a radius as the first filter then apply other heuristics to filter after that. Geo coding raises other issues. For example, some geo coding engines return multiple lat/long pairs (for one address). We deal with only about 2,500 addresses from 4 data sources with quarterly updates and about a 10% turnover rate. We use human review. We set a strict "similarity" level then generate a list of match candidates. Humans review this list and evaluate as a match or non-match. We then loosen the similarity requirement and generate another candidate list (but filter any non-matches found by the first filter). That candidate list is evaluated by humans. Loosen, (rinse) and repeat. After matching comes more fun, merging. Put another way, once you have a match that is not exact, how do you reconcile the differences? Given two differences, which is right? In reply to Re: Efficient Fuzzy Matching Of An Address
by jimX11
|
|