I don't understand half of this, but seems very useful. Need better naming/explanation of use-cases (if I don't understand it, maybe people who are in need of these services don't also).
Record linkage is conceptually pretty simple--it's just deduplicating records (usually of people, in census or medical or commercial or other scenarios) that don't have unique identifiers. It's tough because it's O(N^2) and because you're trying to build a bunch of disparate information into your matching decision.
It's helpful to hear how confusing it is. I stumbled around for literally years trying to build my own solution until finally a kind soul on the Data Science board on Stack Exchange directed me to the Record Linkage Wikipedia page, and I figured out some of the right terms to search for.
I'll try adding some more examples up top and use case scenarios up top.
I'm also hoping to add some info to help differentiate the software a bit more, but unfortunately most of them are barely past the proof-of-concept stage in terms of usability.
3 comments
[ 5.7 ms ] story [ 21.2 ms ] threadI don't understand half of this, but seems very useful. Need better naming/explanation of use-cases (if I don't understand it, maybe people who are in need of these services don't also).
It's helpful to hear how confusing it is. I stumbled around for literally years trying to build my own solution until finally a kind soul on the Data Science board on Stack Exchange directed me to the Record Linkage Wikipedia page, and I figured out some of the right terms to search for.
I'll try adding some more examples up top and use case scenarios up top.
I'm also hoping to add some info to help differentiate the software a bit more, but unfortunately most of them are barely past the proof-of-concept stage in terms of usability.