Sentiment analysis for mobile brands reviews using Convolutional Neural Networks (CNN) / Puteri Ika Shazereen Ibrahim
Sentiment analysis is widely used to mine the web content on web, including texts, blogs, reviews, and comments. It is used to analyse the people opinion. As for product company to monitor their product is by reading all the reviews which is the tedious task and take a lot of time. Sentiment analysi...
Saved in:
Main Author: | |
---|---|
Format: | Thesis |
Language: | English |
Published: |
2020
|
Subjects: | |
Online Access: | https://ir.uitm.edu.my/id/eprint/55497/1/55497.pdf |
Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
Summary: | Sentiment analysis is widely used to mine the web content on web, including texts, blogs, reviews, and comments. It is used to analyse the people opinion. As for product company to monitor their product is by reading all the reviews which is the tedious task and take a lot of time. Sentiment analysis can help to analyze these opinioned data and extract some important insights which will help other users to make the decision. However, because a lack of sufficient labeled data in the field of Natural Language Processing (NLP) it is becoming challenging. To solve that, the CNN approach has been proposed in sentiment analysis because of it learning ability and extract the important feature. The methodology for the project has been designed for CNN implementation that consists of five phases which are preliminary study, data collection, system design, system implementation, and result. The CNN model has shown great accuracy. The parameter tuning has been done to get the great accuracy of the model. The result of the experiment shows that CNN gives high accuracy when the train and test split is 90:10, with 500 input vector, feature map = 200, filter size = 8-12, dropout= 0.1, and batch size = 64. This combination has produced a higher accuracy of the CNN model in this project which is 95%. It is shown that this model needs more input or data to make it more accurate. The evaluation result of this model also has shown that this model is good for sentiment analysis. For future work to improve this project, more data will be used for training and Malay reviews also will be include. After that, real-time data also will be available for users monitoring their products. |
---|