The abstracts published by the Journal of the American Statistical Association in the time span 1946–2016 have been examined in order to identify relevant timings in the recent history of statistics and retrieve past and current topics that have drawn the attention of one of the most influential communities of statisticians in the world. The focus is on clusters of words that, over time, share a similar trajectory of occurrences in the issues of the journal and on the effect of different choices in the number of clusters. When arrangements in coarser and finer groupings have been compared and contrasted, an interesting nested structure has emerged. Moreover, results have highlighted the conjoint effect of word cycle synchrony and word popularity, which are two of the most important features to be accounted for by the researcher in reading the output of a curve clustering based on observations of word frequencies from a chronological perspective. The research also shows that a knowledge-based system (a computer-based system that supports human learning, endowed with a knowledge-base, a statistical learning engine and a user interface) is able to achieve an effective representation of abstracts and that many elements of the history of statistics may be gleaned by reading the abstracts of a large number of papers and considering ‘texts as data’.
