Lexicon modelling for automatic speech recognition of native and non-native speakers