A better fitness measure of a text-document for a given set of keywords

摘要

We present a new fitness measure BW(D) for a text-document D against a set of keywords W. The fitness evaluation forms a basic operation in information retrieval. The measure BW(D) differs from other measures in that it accounts for both the frequency of the keywords and their clustering characteristics. It also satisfies the important properties of monotonicity and super-additivity, which do not hold for either of the well-known Paice-measure and the mixed-max–min measure. We give efficient algorithms for computing BW(D) and a generalized form BαW(D) of it.