BayesWipe cleans your structured database without requiring domain knowledge, a prior clean database, or hiring a dedicated data quality team.
You are welcome to try out the software with your own datasets. Just use a comma-separated file. You can also get some sample data to operate on.
BayesWipe does not need any time consuming expert input. Just provide the database, set an expected percentage of error, and you're done!
BayesWipe is based on Bayes networks and an advanced error model that accurately predicts real-life errors.
There is only one parameter to control: the level of errors you expect in the database.
Supports cleaning data completely offline, in-situ. Creates a copy of the database that is cleaner and highlights changed data.
Acts like a mediator to get answers from data sources not in your control. Learns from a sample of the database, and then finds efficient rewritten queries to quickly retrieve relevant data.
BayesWipe follows probabilistic principles, by finding a probability for every tuple, and for every possible correction; thus finding the best tuple.