Emotion classification based on crime news using SVM machine learning
Emotion may be shown in a variety of manners. These include voice, written texts, and facial expressions and movements. Emotions are divided into six separate categories such as fear, joy, sadness, anger, disgust and surprise. Emotion classification in text, particularly in crime-related news articl...
Saved in:
| Main Author: | |
|---|---|
| Format: | Article |
| Language: | en |
| Published: |
College of Computing, Informatics, and Mathematics
2025
|
| Subjects: | |
| Online Access: | https://ir.uitm.edu.my/id/eprint/127992/1/127992.pdf https://ir.uitm.edu.my/id/eprint/127992/ https://fskmjebat.uitm.edu.my/pcmj/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Emotion may be shown in a variety of manners. These include voice, written texts, and facial expressions and movements. Emotions are divided into six separate categories such as fear, joy, sadness, anger, disgust and surprise. Emotion classification in text, particularly in crime-related news articles, is a crucial task for understanding public sentiment. Classification emotion in normal text is complex, and it becomes even more challenging with news text that does not specify the emotions that they convey, adding complexity to the process. Therefore, in order to solve the problem is to develop an emotion classification system for crime news. The dataset is collected from a reliable news source which are The Star, The Sun and NST. The model is developed using Support Vector Machines (SVM), utilizing a dataset scraped from the news website. The system uses Word2Vec for word embedding to capture semantic relationships and contextual meaning in the text. The methodology follows the Modified Waterfall model, which adapts the traditional Waterfall process while allowing for flexibility and iterative feedback during development. This includes phases of requirement analysis, system design, implementation, testing, and deployment. The project follows a linear sequence but allows for feedback loops and adjustments during key phases like testing and evaluation. The system is evaluated using performance metrics such as accuracy, precision, recall, and F1 score. Preliminary results indicate that the model achieves an accuracy of 81%, precision of 76%, recall of 60%, F1-Score of 62% Word2Vec SkipGram compared to Word2vec CBOW. Future work will focus on the user interface with features like detailed visualizations, interactive dashboards and support for multiple languages to greatly enhance the overall user experience and accessibility. |
|---|