Real-time name mispronunciation detection

A system comprising:

• a processing system;
• a storage system;
• and
• instructions stored on the storage system that when executed by the processing system direct the processing system to at least:
• prior to a start of a meeting, obtain a name of each meeting participant invited to the meeting and an associated user identifier of each meeting participant invited to the meeting;
• generate a custom language model by training the model using the name of each meeting participant invited to the meeting to enable the custom language model to correctly identify and transcribe names of participants spoken during the meeting;
• obtain a reference pronunciation of each meeting participant invited to the meeting from a reference pronunciation resource using the associated user identifier of each meeting participant invited to the meeting;
• receive audio input of a speaker;
• obtain a transcript of the audio input;
• identify a name from text of the transcript based on names of each meeting participant invited to the meeting;
• extract a portion of the audio input corresponding to the name identified from the text of the transcript;
• obtain a reference pronunciation for the name using a user identifier associated with the name;
• with an acoustic signal processor, obtain a pronunciation score for the name based on an electronic comparison between phoneme levels of the reference pronunciation for the name and the portion of the audio input corresponding to the name;
• determine whether the pronunciation score is below a threshold;
• and
• in response to determining the pronunciation score is below the threshold, notify the speaker of a pronunciation error and output a practice experience in which the system records additional audio input of the speaker repeating the name for which the pronunciation score was below the threshold and generating additional pronunciation scores for the additional audio input to allow the speaker to achieve correct pronunciation of the name.

A computer-readable storage medium having instructions stored thereon that, when executed by a processing system, perform a method comprising:

• prior to a start of a meeting, obtaining a name of each meeting participant invited to the meeting and an associated user identifier of each meeting participant invited to the meeting;
• generating a custom language model by training the model using the name of each meeting participant invited to the meeting;
• obtaining a reference pronunciation of each meeting participant invited to the meeting from a reference pronunciation resource using the associated user identifier of each meeting participant;
• storing the reference pronunciation of each meeting participant in a local memory;
• receiving audio input of a speaker;
• obtaining a transcript of the audio input;
• identifying a name from text of the transcript based on names of each meeting participant invited to the meeting;
• extracting a portion of the audio input corresponding to the name identified from the text of the transcript;
• obtaining a reference pronunciation for the name using a user identifier associated with the name;
• with an acoustic signal processor, obtaining a pronunciation score for the name based on an electronic between phoneme levels of the reference pronunciation for the name and the portion of the audio input corresponding to the name;
• determining whether the pronunciation score is below a threshold;
• and
• in response to determining the pronunciation score is below the threshold, notifying the speaker of a pronunciation error and outputting a practice experience in which the system records additional audio input of the speaker repeating the name for which the pronunciation score was below the threshold and generating additional pronunciation scores for the additional audio input to allow the speaker to achieve correct pronunciation of the name.