| Pub. Date | : July, 2023 |
|---|---|
| Product Name | : The IUP Journal of Computer Sciences |
| Product Type | : Article |
| Product Code | : IJCS010723 |
| Author Name | : A C Nanayakkara and G A D M Thennakoon |
| Availability | : YES |
| Subject/Domain | : Management |
| Download Format | : PDF Format |
| No. of Pages | : 16 |
Textual content sentiment analysis is important for a wide range of natural language processing activities. Particularly, the growth of social media stimulates a huge need for sentiment analysis, which is used to extract relevant statistics from massive data on the Internet. The study focused on handling the sentiment analysis problem by employing Deep Learning models with GloVe word embedding, which was motivated by the accomplishments of Deep Learning. A basic Neural Network (NN) was utilized as the experiment's foundation, together with Convolutional NNs (CNNs) and a Long Short-Term Memory (LSTM) NN architecture to categorize the tone of comments left on one of the most popular YouTube videos. The study recommends NN model with LSTM to overcome the shortcomings of the traditional recurrent neural community. In the comparison test between the densely connected neural network model, the CNN model, and the LSTM model, based on training accuracy (81%, 86%, 87%), testing accuracy (64%, 74%, 84%), and over fitting indicators (15, 12, 3), the LSTM model with GloVe word embedding outperformed both the CNN and simple NN. Future studies are encouraged to test different embedding methods with diversified datasets.
Sentiment analysis is a widely used text content category method that analyzes an incoming message and indicates whether or not the crucial opinion is positive or negative. It is an essential tool for understanding how people process textual information and has widespread application in the industry (Aung and Myo, 2017; and Ergul et al., 2021). Although there are several ways to do sentiment analysis (Li et al., 2010; Hangya and Farkas, 2017; and Chen and Chen, 2019), their accuracy and efficiency have been severely hindered by numerous processing issues. In order to deal with these issues, this study focuses on adopting a Deep Learning-based strategy (Recurrent Neural Network with GloVe Word Embedding).
Sentiment analysis, Deep Learning, CNN, LSTM, GloVe word embedding