1Associate Professor, CSE Department, Chandigarh College of Engineering and Technology (Degree Wing), Sector-26, Chandigarh, India
2Student, CSE Department, Chandigarh College of Engineering and Technology (Degree Wing), Sector-26, Chandigarh, India
3Student, CSE Department, Chandigarh College of Engineering and Technology (Degree Wing), Sector-26, Chandigarh, India
(*Corresponding author) email id: varun3dec@gmail.com
This paper reports the results of text classification which is one of the most interesting challenges that has been discovered and researched upon under natural language processing since its inception. GloVe embeddings from Stanford University have been used as the pre-trained weights to the embedding layer and classification task has been performed using the powerful convolution neural networks. The task in hand was to classify the texts based on the training of the model on the popular 20 newsgroup data set. Significant improvement in terms of accuracy has been shown by use of a certain text processing measure that is pickling of the tokeniser which is discussed ahead in the paper. Keras (version 2.0.5) has been used to implement the problem in Python (version 3.5.2) as its deep learning library with TensorFlow (version 1.3.0) as the backend.
NLP, CNN, GloVe embeddings, Newsgroup classification, Deep learning, Machine learning, Artificial intelligence
