Correcting and modelling pronunciation through multimodal resources in speech-language therapy.