A set of segmented and aligned word pairs/triples/..., such as one held in a soundcorrs
object, can be turned into a contingency table in more than one way. Perhaps the simplest option is to see how often various segments from one language correspond to various segments from another language, which is the kind of table this function produces. Correspondences can be reported in absolute or relative numbers, and can represent the number of times the given correspondence occurs, or in how many words it occurs (the same correspondence can occur more than once in a single pair/triple/... of words, e.g. in German koala : French koala, the correspondence G a : Fr a occurs twice). When the numbers are relative, each row in the table adds up to 1. In theory, summary.soundcorrs
can support a soundcorrs
objects with any number of languages in it, but the legibility of the output drops very quickly when that number exceeds two.