Nada Ahmed Marmouch
Mariam Yasser Salah
Rawan Ahmed mohy
Philip George Fayek
Dr. Diaa Salama Abd Elminaam
Eng. Mahmoud ElSahhar
Our lives have become increasingly reliant on social media. It has evolved into one of the most vital information sources. In this work, we propose machine learning and deep learning algorithms. LSTM (one to three layers) and GRU are two deep learning approaches (one to three layers). Six machine learning techniques are used to compare the performance of the proposed methodologies. Decision tree (DT), logistic regression (LR), K nearest neighbor(KNN), random forest (RF), support vector machine (SVM), and NaiveBayes (NB) are the six machine learning approaches.Keras-tuner is used to optimise the parameters of deep learning techniques, whereas a grid search is used to optimise the parameters of machine learning techniques. Three Benchmark datasets were used to train and test models. For the baseline machine learning model and word embedding feature extraction method for deep neural network methods, two feature extraction methods (TF-ID with N-gram) were utilised to extract critical features from the three benchmark datasets. The proposed deep learning techniques always show the best performance because of their ability to learn the discriminatory features through the multiple hidden layers.LSTM(one layer) showed the best cross-validation accuracy (66.79%) on the dataset. In the case of the LSTM(two layers) showed the best cross-validation accuracy (66.07%). Finally, GRU (two layers) showed the best cross-validation accuracy (66.73%). The propsed framework has been categorized into some steps . Experimental results on bench challenging datasets demonstrate that our methods can achieve better performance than numerous state-of-the-art methods.
1-The system shall be reliable and provide suitable performance.
2-Instructors can be able to rely on the system in correcting the students’ answers and get a suitable grade according to it.
3- The system shall provide a high accuracy in less time.
4- Each question shall be graded according to its error percentage whether it exists or not.
1-For subjective questions, auto-correction is seen as a solution to tackle one of academic most known problems.
2-The instructor used to spend a lot of time on their own correcting essays and subjective questions.
3- The aim is to develop a system that uses a certain technique to match synonyms and keywords, then checks the student’s answer to determine the student’s score.
Documents and Presentations
You will find here the documents and presentation for our proposal.
You will find here the documents and presentation for our SRS.
You will find here the documents and presentation for our SDD.
You will find here the documents and presentation for our Thesis
DeepCorrect :Building an Efficient Framework for Auto-Correction for Subjective Questions Using GRU_LSTM Deep Learning