We propose an incremental algorithm for clustering duplicate tuples in large databases, which allows to assign any new tuple t to the cluster containing the database tuples which are most similar to t (and hence are likely to refer to the same real-world entity t is associated with). The core of the approach is a hash-based indexing technique that tends to assign highly similar objects to the same buckets. Empirical evaluation proves that the proposed method allows to gain considerable efficiency improvement over a state-of-art index structure for proximity searches in metric spaces.