Arabic Speech Sentiment Analysis Using Machine Learning: A Case Study of COP28
| dc.contributor.advisor | Professor Khaled Shaalan; Dr Manar Alkhatib | |
| dc.contributor.author | ALMUALLA, SHEIKH ABDULAZIZ | |
| dc.date.accessioned | 2024-08-14T12:24:49Z | |
| dc.date.available | 2024-08-14T12:24:49Z | |
| dc.date.issued | 2024-06 | |
| dc.description.abstract | In recent years, the UAE has played a pivotal role in advancing the global climate agenda by hosting significant events such as the COP28. COP28 served as a crucial platform for international dialogue and cooperation among nations to address climate change and accelerate efforts to mitigate its impacts. In an era characterized by rapid technological advancements, the development of Arabic speech recognition systems emerges as a crucial frontier in enhancing accessibility, efficiency, and usability across various domains. Despite significant advancements in speech recognition for languages like English, challenges persist in adapting these technologies effectively to accommodate the unique characteristics of Arabic. Within this context, exploring Arabic speech recognition within the framework of COP28 serves as a compelling case study. This research integrates speech recognition technologies at COP28 and holds the potential to streamline communication and enhance accessibility for Arabic-speaking delegates and stakeholders. Through a comprehensive investigation of various speech recognition models, including CNN, BI-LSTM, GRU, and hybrid architectures such as CNN-BI-LSTM and GRU-BI-LSTM, valuable insights can be gained into their performance and efficacy within the unique context of Arabic speech recognition. Analysing key metrics such as accuracy across different sentiment categories – positive, negative, and neutral – provides a nuanced understanding of each model's strengths and limitations. The hybrid GRU and BI-LSTM model takes the lead, showcasing outstanding performance with an accuracy rate of 94%. Close behind is the standalone GRU technique, achieving an accuracy of 93%. Subsequently, both the CNN-BI-LSTM and CNN models follow suit with accuracies of 91% and 90%, respectively. The results showed the robustness and the effectiveness of the proposed models. | |
| dc.identifier.other | 22002552 | |
| dc.identifier.uri | https://bspace.buid.ac.ae/handle/1234/2665 | |
| dc.language.iso | en | |
| dc.publisher | The British University in Dubai (BUiD) | |
| dc.subject | Arabic speech recognition, sentiment analysis, deep learning, COP28 | |
| dc.title | Arabic Speech Sentiment Analysis Using Machine Learning: A Case Study of COP28 | |
| dc.type | Dissertation |