reading a set of clusters from a database, each cluster in the clusters including a plurality of records stored in a table of the database, wherein the plurality of records is received from a plurality of independent data sources and wherein the plurality of records is associated with a plurality of entities, wherein at least one record in the plurality of records includes incorrect data;

extracting an oversized cluster in the set of clusters;

performing a breadth-first search (BFS) on the oversized cluster, the BFS generating a list of visited records;

terminating the BFS upon determining that a size of the list of visited records exceeds a maximum size, wherein the size of the list of visited records is less than a size of the oversized cluster;

generating a new cluster from the list of visited records and adding the new cluster to the set of clusters, wherein the new cluster comprises a set of records in the visited records associated with a single entity in the plurality of entities; and

writing the new cluster to the database by updating each of the visited records in the table to identify the new cluster.