Spoken English Intelligibility Remediation with PocketSphinx Alignment and Feature Extraction Improves Substantially over the State of the Art

Automatic speech recognition is used to assess spoken English learner pronunciation based on the authentic intelligibility of the learners’ spoken responses determined from deep neural network (DNN) model predictions of transcription correctness. Using numeric features produced by PocketSphinx alignment mode and many recognition passes searching for the substitution and deletion of each expected phoneme and insertion of unexpected phonemes in sequence, the DNN models achieve 97% agreement with the accuracy of Amazon Mechanical Turk crowdworker transcriptions, up from 75% reported by multiple independent researchers. Using such features with DNN prediction models can help computer-aided pronunciation teaching (CAPT) systems provide intelligibility remediation. We have developed and published free open source software so that others can use these techniques.