On some quality criteria of bipolar linguistic summaries

The quality measures for bipolar linguistic summaries of data, as proposed in our previous work [1], are further developed. The summaries introduced in [2] are assumed to be an extension of the “classical” linguistic summarization (cf. [3], [4]), a human-consistent data mining technique revealing complex patterns present in data. This extension consists in using the “and possibly” to build a summary and introducing the notion of context to determine the validity of the summary. We present a more detailed description of summaries quality measures/criteria and reports results of more extensive computational experiments.