Ranking contemporary American poems

The purpose of this article is to examine the linguistic differences between poems written by 'amateurs' and those written by 'professionals' and then to use these characteristics to rank a number of contemporary American poems. The corpus of poems used consist of 100 poems randomly selected from a recent anthology of professional poets and a control group of 100 poems written by amateurs. The poems were reduced to ninety-eight linguistic and psycholinguistic variables, and these were used in a machine learning algorithm to build an ensemble classifier. The accuracy of the classifier was 84.5%. The probability scores of the individual poems was then used to rank the professional poems on a continuum representing amateur at one extreme and professional at the other, thereby providing an objective means of ranking contemporary poems.