Review on Student Feedback Analysis using Deep Learning
Deokar Anuja T1., Shejwal Asmita S2., Dumbre Ashwini B3., Prof S.S.Gore4
[email protected], [email protected], [email protected], [email protected]
Department of Computer Engineering
Jaihind College of Engineering, Kuran
Feedback plays a key role in improving quality. To ensure improvement in teaching method and facilities provided by college, opinion of the students should be properly analysed and used. Text Sentiment analysis method are used to carry out such analysis. It can be performed in two ways – Machine Learning approach and Lexicon based approach. Presently, the teacher evaluation and feedback analysis are based on identifying student’s opinion. Methods used for such classification are Naive Bayes, Voting ensemble method. Along with determining polarity, classifying feedback as strength, weakness and suggestions can improve to be more beneficial. Success of deep learning inspires us to propose a better and efficient system. The System that will use Word2Vec for text processing, Convolution Neural Network for automatic feature extraction. Supervised Support vector Machine will be used for final classification. Based on classification the text summarization will be done and distributed as strength, weakness and suggestion.
Keywords: Supervised learning, Convolution neural network, feature extraction, Sentiment Analysis, Deep learning
Youth is the hope and future of nations. Today’s youth are students. Students play a vital role in society. It is primary necessity that education given to students needs to be quality education and must consider their views regarding it. Technology has allowed students to explore new fields. Hence considering student point of view to improve educational system can be beneficial. To keep track of performance and help improve abilities of faculty, opinions of students can be helpful. Finding out the subjective meaning from opinions is the major task. Sentiment analysis can be one way to do it. Sentiment analysis is contextual mining of text which identifies and extracts subjective information in source material, and helps to understand the social sentiment when monitoring an organization. Natural language processing (NLP) have to do with building of computational algorithms to automatically analyze and represent human language. The branch of natural language processing is shifting from statistical methods to neural network methods. Deep learning methods are achieving state-of-the-art results on challenging machine learning problems for example describing photos and translating text from one language to another. Our proposed system will be useful in educational sector. The feedback of students, collected in the form of text, will be analyzed using word2vec and CNN. The proposed system will result in classification of feedback as strength, weakness and suggestions to faculty.
Most existing sentiment analysis algorithms were designed for binary classi?cation, meaning that they assign opinions or reviews to bipolar classes such as Positive or Negative. A series of experiments with convolutional neural networks built on top of word2vec are described in 2. The results of experiment show that unsupervised pre-training of word vectors is an important ingredient in deep learning for NLP.In 7 the paragraph of sentences given by the customer is accepted and after extracting each and every word, they are checked with the stored parts of speech, articles and negative words. After checking against the database, Context free Grammar (CFG) is used to validate proper formation of the sentences.
In 3 Automatic evaluation system based on sentiments to overcome drawback of traditional questionnaire system. Feedback is collected in the form of running text and sentiment analysis is performed to identify important aspects along with the orientations using supervised and semi supervised machine learning techniques. It focuses more on subjective sentences and not on objective sentences. The scores are collected and aggregated to calculate final result. Term Frequency – Inverse Document Frequency (TF-IDF) and Naïve Bayes (Unigram, Bigrams) methods is used. It does not use advanced machine learning techniques and so the results were not accurate.Author of 4 states Text Mining techniques are broadly extended to classify the effective improvement of sentiment polarity analysis. Different techniques like Support Vector Machine (SVM), KNN and Decision tree are generally used but they are not always effective. Reducing the feature in data pre-processing stage and teaching sentiment analysis using voting ensemble method of machine learning are proposed and compared with existing typical machine learning for sentiment analysis. The system achieves accuracy improvement of subjective polarity in sentiment analysis. Lack of weight assignment for feature extraction is observed. Methods like Naïve Bayes, ID3, J48 Decision Tree are used. The system described in 6 evaluates faculty and rates them with certain specified parameter to improve academic and education standard. The system is based on attribute and uses multipoint rating system. System uses text mining for deriving high quality information. Academic performance of students is considered when using the feedback given by them. Weights are assigned to feedback based on academic performance and sincerity. Multipoint rating is provided. The number of comparisons can be reduced in the system for more effective system. Naive Bayes method is used for text mining.
In 1 Pre-trained Word2Vec for text pre-processing and to gain vector representations of words which will be the input for suitable Convolutional Neural Network (CNN) architecture for deep features extraction is applied. Rectified Linear Unit and Dropout functions is used to improve the accuracy. Support Vector Machine classifier was used to predict the final classification. Author of 9 combines the advantages of CNNs and SVM, and constructs a text sentiment analysis model based on CNNs and SVM. The pretrained word vector is used as input, and CNNs is used as an automatic feature learner, and SVM is the final text classifier. It is found that the accuracy of using CNN model results in better other models of depth learning, which shows that CNN model is more suitable to deal with text affective classification problem 8.
Natural Language Processing:
Natural Language Processing (NLP) is science of deriving meaning from natural language. To predict meaning of natural language initially rule based system was used. But one cannot write rule for everything in natural language. Increase in rules resulted in increase in their conflicts. After rule-based system Statistical Machine Learning like naive bayes proved to give better results.
Problems for NLP when using statistical machine learning:
Deriving meaning of natural language is not that easier. Consider following example: “It is a good product.” The sentiment of this sentence can directly be analyzed as positive word ‘good’. But in “I didn’t receive service on time.” Negation is used and receive is neither positive nor negative. Some rare words like ‘shoddy’ are not known to classifiers. Analyzing the misspelled words like gr8 for greater proves to be difficult. Sarcastic sentence like “Some cause happiness wherever they go; others whenever they go.” have negative meaning with positive words.
Deep Learning is machine learning technique that learns features and tasks directly from the data. Data can be images, text or sound. The term deep usually refers to the number of hidden layers in neural network. Deep learning has come to hype post 2012. Access to large datasets and availability of powerful computational machines has paved a way to emergence to deep learning. Deep Learning has amazing results in natural language processing. Internet now a day is huge source data as a result of growing number of users. Large amount of data on internet is in natural language. Deep learning in NLP has its application in customer support, summarization, research, sentiment analysis.
Word2Vec is the name given to a class of neural network models that, given an unlabelled training corpus, produce a vector for each word in the corpus that encodes its semantic information
Fig: Word2Vec Algorithms 10
Main idea behind Word2Vec is to predict between every word and its context word. Two algorithms are used in word2vec.
Skipgram: Used to predict context words from given target word.
Continuous Bag of Words (CBOW): Predict target words from bag of words context
Two training methods are used
Convolution neural network can extract meaningful feature representation from input samples effectively, but the classification ability of fully connected classification layer 9 is weak for nonlinear separable data. The pretrained word vector is used as input, and CNNs is used as an automatic feature learner. convolutional neural networks (CNN) trained on top of pre-trained word vectors for sentence-level classification tasks. The most prominent feature of CNN is local sensing and parameter sharing 8.
Convolutional neural network has been able to enhance the machine learning system is mainly the use of three important concepts: sparse connection, parameter sharing, equivalent representation. Convolution neural network is a kind of artificial neural network, which is a simple neural network constructed by using convolution operation instead of matrix multiplication operation of at least one layer in neural network. Convolutional neural networks (CNN) utilize layers with convolving filters that are applied to local features 2.
Fig: Convolution Neural network 8
SVM is a supervised machine learning model, which is two-classification model., SVM try to find the best compromise between the complexity and the learning ability, in order to get the best generalization ability. Among them, the learning ability refers to the ability to detect any error free samples, and the complexity of the model refers to the learning that the model can get from a specific training sample. SVM represents the data feature vectors in the feature space, the support vector refers to those samples from the training data that are closest to the classified hyper plane. For linear non-separable data, SVM can map the data into a high dimensional space through a kernel function, and then transform the linear non-separable problem into a linear separable problem 9.
Students will give the feedback in the form of text and opinions regarding the faculty members and facilities provided by educational institute. The system will make use of Word2Vec for text processing. In word2vec, a distributed representation of a word is used. Take a vector with several hundred dimensions (say 1000). Each word is represented by a distribution of weights across those elements. So instead of a one-to-one mapping between an element in the vector and a word, the representation of a word is spread across all of the elements in the vector, and each element in the vector contributes to the definition of many words. Convolution Neural Network will the perform automatic feature extraction. Supervised Support Vector Machine will be used for final classification. Finally, the faculties will be notified about their strength and weakness and suggestion given by students.
Fig: Proposed System ArchitectureConclusion
Students Feedback analysis using deep learning can be useful in educational sector to find strength, weakness and suggestion of faculty.
We are thankful to Jaihind College of Engineering for providing the resources. We are also thankful to Prof. Khatri A.A, project co-ordinator, and Prof. Gore S.S, Guide, for continuous support
A bubakr H. Om babi, O nsa Lazeez, Wael Ouarda, Adel M. Alimi “Deep Learning Framework based on Word2Vec and CNN for Users Interests Classification” 978-1-5386-0667-4/17/$31.00 ©2017
Yoon Kim “Convolutional Neural Networks for Sentence Classification” Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 1746–1751,October 25-29, 2014, Doha, Qatar. ©2014
Alok Kumar, Renu Jain “Sentiment Analysis and Feedback Evaluation” 978-1-4673-6747-9/15/$31.00 2015
Chakrit Pong-Inwong, Konpusit Kaewmak “Improved Sentiment Analysis for Teaching Evaluation Using Feature Selection and Voting Ensemble Learning Integration” 978-1-4673-9026-2/16/$31.00 ©2016
Neelima Bhatia, Arunima Jaiswal “Automatic Text Summarization and it’s Methods- A Review” 978-1-4673-8203-8/16/$31.00 c 2016
Krishnaveni K S, Rohit R Pai, Vignesh Iyer “Faculty Rating System Based on Student Feedbacks Using Sentimental Analysis” 978-1-5090-6367-3/17/$31.00 ©2017
Biswarup Nandi, Mousumi Ghanti, Souvik Paul “TEXT BASED SENTIMENT ANALYSIS” 978-1-5386-4031-9/17/$31.00 ©2017
Fan Xia, Zhi Zhang “Study of Text Emotion Analysis Based on Deep Learning”978-1-5386-3758-6/18/$31.00 c 2018
Yuling Chen, Zhi Zhang “Research on text sentiment analysis based on CNNs and SVM” 978-1-5386-3758-6/18/$31.00 c 2018
Tomas Mikolov, Kai Chen, Greg Corrado, and Je_rey Dean. E_cient estimation of word representations in vector space. 01 2013.