Electronic learning system

A method, comprising:

• receiving, by an electronic communications system, an electronic assessment of a user, wherein the electronic communications system includes memory and a processor;
• reducing, by the electronic communications system, noise in audio features of the electronic assessment, wherein the reducing the noise is based on a sample audio spectrogram of the user from previous recordings;
• receiving, by the electronic communications system, historical electronic information;
• analyzing by the electronic communications system the electronic assessment based on the historical information, wherein the analyzing the electronic assessment includes:
• analyzing sound wave slots in a time T, wherein a portion of the sound wave slots with a null frequency,
• determining that the portion of the sound wave slots with null frequency do not affect speech classification,
• removing the sound wave slots with the null frequency,
• segmenting other sound wave slots with frequency into smaller audio clips based on a set timespan, wherein the smaller audio clips result in an equal length input for a particular question asked of the user,
• wherein the equal length of each of the smaller audio clips is based on a total number of system generated communications sent to the user within a particular amount of time,
• normalizing audio volume of the electronic assessment to be the same as other electronic assessments from other users,
• padding each of the smaller audio clips, wherein the padding is added to the smaller audio clips' starts and ends,
• wherein the padding maintains an equal wave input for each of the smaller audio clips to prevent the smaller audio clips from having different lengths, and
• wherein the padding of each of the smaller audio clips does not affect the classification of each of the smaller audio clips,
• excluding any of the smaller audio clips that are labeled as including no answer;
• and
• determining, by the electronic system, a learning disability of the user based on the analyzing the historical information and the electronic assessment, wherein the determining the learning disability is based on:
• a word error rate (WER) and a character error rate (CER);
• and
• receiving, by the electronic communication system, a training set,
• labeling the training set as:
• no answer, wrong, partially correct, or correct,
• conducting, by the electronic communication system, training of the electronic communication system, wherein the training includes running a Long Short-Term Memory (LSTM) algorithm and a convolutional neural network (CNN) algorithm,
• wherein the LTSM algorithm is configured to apply a reading text task to maintain the dependencies between words as a sentence and overcome corrections by other user;
• classifying electronic inputs from sampled student recorded answers in order to be classified into letters and converting to text to be compared with the question text and result in one of the following categories:
• correct, partially correct, or wrong;
• combining the CNN algorithm results and the LSTM algorithm results;
• and
• validating the combination of the outputs of the CNN algorithm and the LSTM algorithm, wherein the validating is used to determine the word error rate (WER) and the character error rate (CER).