"For the fifty-eleventh time" : Examining cross-linguistic properties of hyperbolic numerals and quasi-numeral expressions through parallel text extraction

Abstract: In some languages, vague and exaggerated quantities can be represented using certain conventionalised numeral expressions with cross-linguistically varying values, such as Danish hundredesytten "117". Hyperbolic quantities can also be expressed using other quantifier expressions (such as English zillion) which, while they do not denote a specific numerical value, have both structural and functional similarities with exact numerals.These hyperbolic numerals and quasi-numerals are common in a variety of primarily informal contexts, but have yet to be subject to a systematic cross-linguistic investigation. In this study, hyperbolic numerals and quasi-numeral expressions are extracted from a massively parallel corpus of film and TV subtitles in a convenience sample of 50 languages, using automatic word alignments and seed expressions in 5 languages. Additional expressions are also obtained through elicitation. The collected expressions are subsequently analyzed and categorized to investigate patterns in distribution, value, morphology, function and usage. Findings include a cross-linguistic preference toward certain powers of the numeral base for hyperbolic numerals, and consistent patterns of construction with varying degrees of cross-linguistic prevalence for quasi-numerals.