A Multi-Dataset Comparative Analysis of Keras and GloVe-Based BiLSTM Models for Sentiment Analysis
Naveen Kumar, Anupam Agrawal · 2023
The form of text on the web expresses feelings, thoughts, and opinions about entities such as brands, products, and people, generating a vast amount of data. These sentiments have a significant impact on businesses and market research. This paper presents an investigation into the performance of the a deep NN in the domain of Natural Language Processing (NLP). Specifically, we implement the BiLSTM model using two different kinds of word embeddings, namely Keras and Global Vectors for Word Representation (GloVe). We further assess the model’s performance on various datasets, which include the airline industry, a crisis scenario, a political party, and a technology company, among others, using various performance metrics in order to show the significance of different embeddings.