Supervised prediction of production patterns using machine learning algorithms

Abstract

When an English word ending in a stop is adapted to Korean, a vowel is variably inserted after the final stop: some words always take the epenthetic vowel, and some never do, while some vary between these alternatives. Although there are different linguistic factors that possibly affect this insertion, it is not easy to determine which pattern will be chosen if a new word comes into the borrowing language. This study conducted classification data analyses of production patterns based on machine learning algorithms including support vector machines and random forests. These two classifiers show similar results where vowel tenseness is the best predictor among all the possible predictors. This indicates that vowel tenseness is most influential in classifying the patterns (no vowel insertion, optional vowel insertion, or vowel insertion). Results suggest that while vowel tenseness remains significant, other factors such as stop voicing and stop place also hold some importance, albeit to a lesser degree. The contribution of this study is that it provides insight into the factors that regulate vowel insertion, and these findings support the need for a behavioral experiment to see if the current results can make right predictions with respect to the behavior of nonce items.