Discussion of word frequency and keywords in text focuses on a method to estimate automatically the stability classes that indicate a word's popularity with time-series variations based on the frequency change in past electronic text data. Compares the evaluation of decision tree stability class results with manual classification results. (Author/LRW)