LARGE VOCABULARY CONTINUOUS SPEECH RECOGNITION FOR VIETNAMESE, A UNDER-RESOURCED LANGUAGE

Hong Nguyen (1, 2) , Pascal Nocera (1) , Eric Castelli (2)

1 LIA - Laboratoire Informatique d'Avignon
2 MICA - International Research Institute MICA

This paper proposes a method to build a Vietnamese Large Vocabulary Continuous Speech Recognition system (Vietna-mese LVCSR system). The difference between Vietnamese and European languages is analyzed and used to adapt a LVCSR system for European languages to Vietnamese. Experiments are implemented on the VNSPEECHCORPUS. The results show that the accuracy of Vietnamese recognition system is increased by using Vietnamese language characteristics .

Mots clés: Index Terms— Automatic speech recognition; Vietna-mese language; under-resourced language; tone recognition; compound noun

Domaines: Informatique [cs]

Hong Nguyen, Pascal Nocera, Eric Castelli. LARGE VOCABULARY CONTINUOUS SPEECH RECOGNITION FOR VIETNAMESE, A UNDER-RESOURCED LANGUAGE. The first International Workshop on Spoken Languages Technologies for Under-resourced languages (SLTU - 2008), 2008, Hanoi, Vietnam. ⟨hal-01311389⟩