Phonetic-context mapping in language identification

This paper deals with the problem of exploiting information from a wide phonetic context for the purpose of language identi cation. Two approaches to language modeling are presented here: 1) modi ed bigrams with a context-mapping matrix and 2) language models based on binary decision trees. Both models were incorporated in a phonotactic language identi er with a double-bigram decoding architecture and were shown to consistently improve the performance of standard bigrams. Measured on the NIST'95 evaluation set, the described system outperforms the state-of-the-art phonotactic components and is, at the same time, computationally less expensive.