Computationally-assisted analysis of early Tahitian oral poetry

Abstract

A computationally-assisted analysis was undertaken of Tahitian oral poetry transcribed in the early 19th century, with the aim of discovering its poetic organization. An automated pattern detection process attempted to recognize many of the organizational possibilities for poetry that have been documented in the literature, as well as be open to unanticipated varieties. Candidate patterns generated were subjected to several rounds of manual review. Some tasks that would have proved difficult to automate, such as the detection of semantic parallelism, were pursued fully manually. Two distinct varieties of meter were encountered: A syllabic counting meter based upon a colon line, and a much less common word stress counting meter based upon a colon line or a list item. The use of each meter was ubiquitous in the corpus, but somewhat sporadic. Word stress counting meter was typically applied to lists, and generally co-occurred with patterns of syllabic counting meter; perhaps in order to enhance metrical effect through an addition of rhythm. For both meters, counts were regulated by an external pattern, wherein they were observed to repeat, increment, form inverted structures, or group into alternating sequences. There appeared to be few limitations as to the possibilities for a pattern‟s starting count or length. Patterns were found to juxtapose freely, as well as alongside unpatterned counts. According to Nigel Fabb and Morris Halle, syllabic counting meter is only otherwise encountered in a style of Hebrew poetry from the Old Testament (Fabb and Halle 2008:268, 271, 283). Word stress counting meter may be unique to Tahitian poetry. The colon also functioned as poetic line for purposes of sound parallelism, which manifested itself in patterns of simple assonance, simple consonance, and complex patterns that combined simpler ones of assonance, consonance, and parallel strings of phonemes. Although sound patterns most often spanned lines, they were sometimes constrained to within a line. Occasionally, they were arranged into inverted structures, somewhat analogous to those noted for counting meter. Some sound patterns were contained within names and epithets, and perhaps served as recurring islands of parallelism. Syntactic parallelism was common, especially in the organization of lists. Occasionally, its application was suggestive of canonical parallelism. Items of syntactic frame lists were often arranged so as to assist patterns of counting meter. A syntactic frame‟s variable elements often belonged to a single semantic category for which there seemed to be no restriction, and which could represent any taxonomic level. There appeared to be complete freedom in regards to the arrangement of syntactic frame patterns, and it was common for several to follow one another in unbroken succession. There is evidence that some of the corpus poetry was memorized. Other evidence suggests that a capacity existed, and perhaps continues to exist, of poetic composition-in-performance.