Systems and methods for comprehensive Chinese speech scoring and diagnosis

A method comprising:

• receiving, with a server, both of a transcript in a tonal language and an audio recording of a person reciting the transcript, the transcript including a number of characters of the language, the server including a processor and non-transitory memory storing computer-readable instructions that when executed by the processor cause the server to perform the steps of the method;
• segmenting the transcript into a segmented text including the characters of the transcript and spaces;
• converting the segmented text to a segmented pronunciation output including a first file having, for each character of the transcript, one or more pinyins without tone markings and a second file having, for each character of the transcript, the same pinyins with associated tone markings;
• determining phoneme boundaries within the audio recording to identify successive individual phonemes between the boundaries, and for each of these individual phonemes identifying a corresponding time-delineated section of the audio recording;
• for each pinyin of the first file, matching each phoneme thereof a corresponding time-delineated section of the audio recording;
• for each character of the transcript, determine the corresponding one or more pinyins of the second file, and determine a tone for the one or more time-delineated sections of the audio recording corresponding to the one or more pinyins of the second file;
• determining a goodness of pronunciation score for each phoneme of each pinyin of the first file by comparing that phoneme to the phoneme in the corresponding time-delineated section of the audio recording;
• determining a correctness of tone score for each character of the transcript by comparing the tone marking for the one or more pinyins for that character in the second file to the tone determined for the corresponding one or more time-delineated sections of the audio recording;
• and
• determining a word-level score for each word in the transcript, the word-level scores being based on the goodness of pronunciation scores and correctness of tone scores.