Zingg learns two models from the data.
One fundamental problem with scaling data mastering is that the number of comparisons increases quadratically as the number of input records increases.
Data Mastering At Scale
Zingg learns a clustering/blocking model which indexes near similar records. This means that Zingg does not compare every record with every other record. Typical Zingg comparisons are 0.05-1% of the possible problem space.
The similarity model helps Zingg to predict which record pairs match. The similarity is run only on records within the same block/cluster to scale the problem to larger datasets. The similarity model is a classifier that predicts the similarity of records that are not exactly the same but could belong together.
Fuzzy matching comparisons
To build these models, training data is needed. Zingg comes with an interactive learner to rapidly build training sets.
Shows records and asks user to mark yes, no, cant say on the cli.