Predicting Potential Difficulties in Second Language Lexical Tone Learning with Support Vector Machine Models

摘要

Second language speech learning is affected by learners' native language backgrounds. Teachers can facilitate learning by tailoring their pedagogy to cater for unique difficulties induced by native language interference. The present study employed Support Vector Machine (SVM) models to simulate how naive listeners of diverse tone languages will assimilate non-native lexical tone categories into their native categories. Based on these simulated assimilation patterns and extrapolating basic principles from the Perceptual Assimilation Model (Best 1995), we predicted potential learning difficulties for each group. The results offer teachers guidance concerning which tone(s) to emphasize when instructing students from particular language backgrounds.