Matching lists of addresses is an increasingly common taskexecuted by business and governments alike. However, due to securityissues, this task cannot always be performed using cloud computing.Moreover, addresses can arrive with spelling errors that can cause non-matches or ‘false negatives’ to occur. Our proposed framework, Post-Match, provides a locally-executed method for address-matching thatcombines the open-source ‘Libpostal’ address-parsing library with our‘postparse’ post-processor code and machine-learning. PostMatch pro-vides improved parsing accuracy compared with Libpostal alone, ap-proaching 96.9%. The matching process features the Jaro-Winkler editdistance algorithm together with XGBoost machine-learning to achievevery high accuracy on public data. PostMatch is open-source (GPL3 li-censed) and available as R script code on Github.
|Title of host publication||Communications in Computer and Information Science|
|Number of pages||15|
|Publication status||Accepted/In press - 05 Oct 2021|