Methods for the automated generation of speech sample asset production scores for users of a distributed language learning system, automated accent recognition and quantification and improved speech recognition

A method for improved speech recognition of non-native speaker speech, the method comprising:

• for a plurality of speech sample assets comprising native speech sample assets;
• sending the native speech sample assets to a plurality of non-native speaker users to perform production exercises;
• receiving production speech sample assets generated by the production exercises;
• sending the production speech sample assets to a plurality of users to perform perception exercises;
• receiving production scores generated by the perception exercises;
• training a supervised machine learning module comprising:
• generating a training set comprising:
• the production speech sample assets,
• the associated production scores;
• and
• associated user background information of the plurality of non-native speaker users;
• and
• training the supervised machine learning module utilising the training set to generate a trained supervised machine learning module;
• and
• for a candidate production speech sample asset of a non-native speaker user:
• inputting the candidate production speech sample asset into the trained supervised machine learning module to recognise user background information of the non-native speaker user;
• categorising the user background information into an accent type category;
• selecting an acoustic model for a speech recognition module in accordance with the accent type category;
• and
• recognising speech of the candidate production speech sample asset using the speech recognition module.