Ratiosoften

Ratiosoften is a term used in corpus linguistics to describe a metric that quantifies how often ratio-related expressions appear in a text collection. The concept is used to study the prevalence of quantitative language across genres, domains, and languages, and to compare how writers communicate proportional relationships in different contexts. The word is a portmanteau of ratio and often, and is described in methodological discussions as a way to summarize a text corpus’s density of ratio language.

Definition and scope: Ratiosoften considers a predefined lexicon of ratio-related forms, including explicit expressions such as “A to B,” “A:B,” “percent,” “ratio of A to B,” and “twice as many,” as well as explicit numeric patterns like “3:1” or “three to one.” It may exclude non-quantitative uses of numerals. A token is counted whenever it matches any item in the lexicon or its morphological variants after standard tokenization.

Calculation: For each document d, let R_d be the number of ratio-expressing tokens and T_d the total tokens. One common formulation is r_d = R_d / T_d. Ratiosoften is then the average over the corpus: ratiosoften = (1/N) Σ_d r_d, where N is the number of documents. In some studies, the measure is scaled to per-million tokens or computed at the sentence level.

Applications and limitations: Applications include comparing textual corpora, studying linguistic style, modeling information density, and tracking changes over time or across languages. Limitations arise from lexicon coverage, tokenization differences, polysemy, and cross-domain normalization needs. See also: frequency analysis, phrase frequency, linguistic statistics.