Tools
Tools
Habib, Kamran (2023) Leveraging Multimodal Data Fusion for Improved Emotion Detection System. Masters thesis, Dublin, National College of Ireland.
Preview |
PDF (Master of Science)
Download (4MB) | Preview |
Preview |
PDF (Configuration Manual)
Download (2MB) | Preview |
Abstract
There is availability of massive data in this digital age and their major source have been social media. They can be in the form of text, images, audio and video. Emotion detection from them have been studied drastically but their merge have not been studied so much. This study with the help of AI can be of great help in detecting emotion from text-images. With the goal to evaluate models for emotion analysis using text, visuals, and their combination, the study performs a number of case studies. It highlights how much better multimodal data is at capturing emotions than unimodal methods. In order to accurately identify emotions, the study analyzes various models and their performances, emphasizing the importance of feature extraction, model selection, and data preparation. The outcomes highlight the potential of AI in improving emotion analysis by demonstrating the success of innovative techniques like convolutional neural networks in interpreting complicated emotional expressions. After the evaluation it was found that multimodal analysis was more successful than unimodal as their Naive Bayes, SVM, Random Forest, KNN and ANN model perform way better than then best model unimodal analysis at 0.87, 0.97, 0.97, 0.97 and 0.98 respectively while best for unimodal was of CNN at 0.97 and SVM at 0.83 for text and for only images best was of CNN at 0.57. This demonstrates the efficacy of text-image fusion in emotion analysis, highlighting the potential of AI in this field.
Actions (login required)
 |
View Item |