Abstract:
" Administrators of tuition centers are currently dealing with
several management issues in addition to students and teachers. One of most time-consuming part
is answers grading process in a tuition class. Some classes have up to 1500 + students in urban
area private tutoring centers. So, teachers need to allocate more time to answer grading. Most of
teachers in private tutoring classes are also government schoolteachers. So, they can’t manage
answers grading process properly.
Automation of the private tutoring sector using machine learning is done in order to address these
problems. The ""TutorHome"" system has been introduced as a remedy. In the core part named
automatic grading functions to remove punctuations, tokenize, and lemmatize text, remove
stopwords, compute cosine similarity between two sentences using Universal Sentence Encoder
and Word2Vec, compute the length of sentences and words, and measure the overlap of a student's
answer with the question prompt. It also defines a function to load a pre-trained Doc2Vec model
and compute the cosine similarity between two documents. These functions are then used to extract
features from a dataset of student answers, which are then used to train and evaluate the machine
learning model to predict the correctness of those answers."
