Accent identification

Abstract

Foreign accent identification is a new challenging problem closely related to other relatively recent fields of the multilingual area such as dialect identification and language identification. The paper describes an automatic identification system for English accents from 6 different European countries. The approach is based on a parallel set of ergodic nets with context independent HMM units. The ergodic topology was also substituted by pronunciation transcription constraints in order to integrate accent specific automatic word recognisers. Considering the complexity of the task, the results can be considered encouraging for further research.