Comparative Analysis of Using Word Embedding in Deep Learning for Text Classification

Abstract

A group of theory-driven computing techniques known as natural language processing (NLP) are used to interpret and represent human discourse automatically. From part-of-speech (POS) parsing and tagging to machine translation and dialogue systems, NLP enables computers to carry out various natural language-related activities at all levels. In this research, we compared word embedding techniques FastText and GloVe, which are used for text representation. This study aims to evaluate and compare the effectiveness of word embedding in text classification using LSTM (Long Short-Term Memory). The research stages start with dataset collection, pre-processing, word embedding, split data, and the last is deep learning techniques. According to the results of the experiments, when compared to the glove technique it seems that FastText is superior, the accuracy obtained reaches 90%. The number of epochs did not significantly improve the accuracy of the LSTM model with GloVe and FastText. It can be concluded that the FastText word embedding technique is superior to the GloVe technique.