In Data Exploration platforms, diversification has become an essential method for extracting representative data, which provide users with a concise and meaningful view of the results to their queries. However, the benefits of diversification are achieved at the expense of an additional cost for the post-processing of query results. For high dimensional large result sets, the cost of diversification is further escalated due to massive distance computations required to evaluate the similarity between results. To address that challenge, in this paper we propose the Progressive Data Diversification (pDiverse) scheme. The main idea underlying pDiverse is to utilize partial distance computation to reduce the amount of processed data. Our extensive experimental results on both synthetic and real data sets show that our proposed scheme outperforms existing diversification methods in terms of both I/O and CPU costs.