No abstract

Basically, we are proposing an automatic segmentation of the vocal signal, in order to tackle the leamer's output assessment problem (As far a non-native pronunciation is concerned rignt now, we are working on samples from leamers of French as a Foreign Language). The approach is underpinned by observations, collected within the instructional realm, and wich point the adavantages of assessing pronunciation at several levels (phonetic features, phonemes). In this paper, we are analyzing the ouptuts of the segmenting process on the (learner-produced) items as well as on corresponding (teacher-produced) ones and compare them.

No abstract

No abstract