Computer assisted linguistic training including machine learning

A system to train a user to speak with a target accent, which comprises:

• a computer system comprising a computer readable storage medium having program instructions embodied therewith, the program instructions executable by a computer to cause the computer to:
• determine improvable aspects of diagnostic input, received at the computer, the diagnostic input representing a user speaking diagnostic phrases with a target accent;
• select from a corpus of reference audio phrases received at the computer from a reference audio source, a focus phrase characterized by at least one of said improvable aspects, said focus phrase being selected, in accordance at least in part, with a learnability score calculated, by said computer, using a deep learning ML model trained to assess, for the user, learning difficulty of the improvable aspects;
• receiving from an input source, performance input representing the user speaking the focus phrase attempting to use the target accent;
• convert the performance input into augmented focus output having audio attributes of a baseline voice of the user with the target accent applied thereby generating an accent-corrected version of the baseline voice, and present the augmented focus output to the user;
• responsive to said presentation, receiving from an input source revised performance input representing the user attempting to replicate the augmented focus output, and determining by said computer, teachable aspects of said revised performance input;
• convert selected teachable aspects of the revised performance input into augmented teaching output having at least one audio attribute of the baseline voice of the user with the target accent applied, and present the augmented teaching output to the user;
• generate similarity scores of the revised performance input based on the revised performance input matching the augmented focus output;
• and
• determine when the user is proficient at speaking the focus phrase based on the similarity scores meeting a proficiency threshold for similarity.

A computer program product to train a user to speak with a target accent, the computer program product comprising a computer readable storage medium having program instructions embodied therewith, the program instructions executable by a computer to cause the computer to:

• determine, using a computer, improvable aspects of diagnostic input, received at the computer, the diagnostic input representing a user speaking diagnostic phrases with a target accent;
• select, using the computer, from a corpus of reference audio phrases received at the computer from a reference audio source, a focus phrase characterized by at least one of said improvable aspects, said focus phrase being selected, in accordance at least in part, with a learnability score calculated, by said computer, using a deep learning ML model trained to assess, for the user, learning difficulty of the improvable aspects;
• receiving from an input source, using the computer, performance representing of the user speaking the focus phrase attempting to use the target accent;
• convert, using the computer, the performance input into augmented focus output having audio attributes of a baseline voice of the user with the target accent applied thereby generating an accent-corrected version of the baseline voice, and present the augmented focus output to the user;
• responsive to said presentation, receiving from an input source revised performance input representing the user attempting to replicate the augmented focus output, and determining by said computer, teachable aspects of said revised performance input;
• convert, using the computer, selected teachable aspects of the revised performance input into augmented teaching output having at least one audio attribute of the baseline voice of the user with the target accent applied, and presenting the augmented teaching output to the user;
• generate similarity scores of the revised performance input based on the revised performance input matching the augmented focus output;
• and
• determine when the user is proficient at speaking the focus phrase based on the similarity scores meeting a proficiency threshold for similarity.