Completely agree. In practice it is never deterministic or probabilistic, it is both. What gets interesting is how those signals interact. Deterministic matches are not always absolute, and probabilistic signals are not independent. A shared phone number might be high confidence in one context and low in another, depending on the surrounding relationships. At some point it stops being about combining techniques and becomes a context problem. You are not just matching attributes, you are evaluating how signals reinforce or contradict each other across a network of records.
Yes thats right, context varies and hence idr platforms should make the least amount of assumptions and let the user make that choice, something we have worked very hard towards at Zingg.
Completely agree. In practice it is never deterministic or probabilistic, it is both. What gets interesting is how those signals interact. Deterministic matches are not always absolute, and probabilistic signals are not independent. A shared phone number might be high confidence in one context and low in another, depending on the surrounding relationships. At some point it stops being about combining techniques and becomes a context problem. You are not just matching attributes, you are evaluating how signals reinforce or contradict each other across a network of records.
That is where the real complexity shows up.
Yes thats right, context varies and hence idr platforms should make the least amount of assumptions and let the user make that choice, something we have worked very hard towards at Zingg.