Detecting loan words computationally
Liqin Zhang, Franz Manni, Ray Fabri, John A. Nerbonne · Contact language library · 2021
Abstract A loanword is a word that is borrowed from one language and adopted into another; examples are the English words toboggan, skunk, and hickory , all of which were borrowed from Algonquian languages. Among languages that are not (closely) related, loan words are recognizable because they are semantically related and are more similar in pronunciation than one would expect by coincidence. This chapter applies techniques for measuring pronunciation similarity, focusing on edit-distance measures and a sound-class based method. The novel issue in loan-word detection is the circumstance that loan words are normally modified to fit the phonology of the borrowing language, meaning that sensitivity in measuring pronunciation similarity may be deprecated.