Automatic reading tutoring with parallel polarized language modeling

A method comprising:

• training a target language model specific to a limited domain of target texts;
• training a garbage language model based on a general dictation grammar;
• providing the trained target language model and the trained garbage language model for use in an automatic reading tutoring system that is configured to:
• display portions of the target texts;
• receive acoustic inputs through a configuration consistent with a user reading the displayed portions of the target texts;
• model the acoustic inputs with reference to the target language model and the garbage language model to indicate whether the acoustic inputs comprise proper readings or miscues relative to the target texts;
• and
• provide feedback via a user output based on whether the acoustic inputs are indicated to comprise proper readings or miscues relative to the target texts.

A medium comprising instructions that are readable and executable by a computing system, wherein the instructions configure the computing system to execute an automatic reading tutoring application comprising a series of iterations of tutoring a user in reading a subject language, wherein each of the iterations comprises:

• providing a text sample comprising target words via a user-perceptible output;
• calculating a language model score for the target words;
• receiving an acoustic signal via a user input after the text sample is provided;
• calculating an acoustic score for the target words with reference to the acoustic signal;
• evaluating whether the acoustic signal comprises a miscue with reference to the text sample based on a weighted comparison of the acoustic score and language model score of the target words with an acoustic score and language model score of a set of garbage words;
• and
• providing user-perceptible feedback based on the evaluation of whether the acoustic signal comprises a miscue.