AFAAN OROMO TEXT SEMANTIC NETWORK ANALYSIS MODEL FOR CLASSIFICATION OF TEXT USING DEEP LEARNING APPROACH

Major Advisor: Dr. M. Kumarasamy (PhD) Natural Language Processing is the intersection of computer science, linguistics and machine learning. The field focuses on communication between computers and humans in natural language and NLP is all about making computers to understand and generate human language. Semantics is a branch of linguistics, aims to study the meaning in language. Semantic Analysis is the branch of natural language processing that aims to understand the meaning of a text. Semantic analysis gives the exact meaning or the dictionary meaning from structures created by syntactic analysis. Semantic networks are becoming more and more popular issue current day. In Semantic networks, we can represent our knowledge in the form of graphical networks. This network consists of nodes representing objects and arcs or edges which describe the relationship between those objects. Deep learning is part of machine learning that can utilize either unsupervised or supervised algorithms. Deep learning is a set of algorithms and techniques inspired by how the human brain works. The main deep learning architectures that used in text classification are simple neural network, convolutional neural network and recurrent neural network. From those deep learning algorithm we selected RNN for the development of prototype. In RNN, the neural network considers the information of previous nodes in a very sophisticated method which allows for better semantic analysis of the structures in the dataset. We developed the LSTM model to train our data for semantic network analysis model. LSTM is a special type of RNN that addresses problems by preserving long term dependency more expertly in comparison to the basic RNN. RNN mostly works by using LSTM or GRU for text classification. The model is evaluated using accuracy and loss by training, testing, and validation dataset and result were recorded and presented by taking differenent number of neuron, number of epoch and batch size alternatively. The final result shows that the model performs finally that we got 97.75% training accuracy, 96.60% testing accuracy, with 0.3185 and 0.7710 loss during training and validation respectively at the end of 30 epoch by using batch size = 2. Finally we got 98.22% training accuracy, 97.08% testing accuracy, with 0.1752 and 0.4278 loss during training and validation respectively at the end of 30 epoch by using batch size = 4. Keywords: Semantics, Semantic analysis, Semantic network analysis, Deep learning, Word embedding, Semantic similarity, Word2vec, modeling, classification