Arabic Sentiment Analysis with Noisy Deep Explainable Model
| dc.contributor.author | Atabuzzaman, Md. | en |
| dc.contributor.author | Shajalal, Md | en |
| dc.contributor.author | Baby, Maksuda Bilkis | en |
| dc.contributor.author | Boden, Alexander | en |
| dc.date.accessioned | 2024-04-01T17:00:07Z | en |
| dc.date.available | 2024-04-01T17:00:07Z | en |
| dc.date.issued | 2023-12-15 | en |
| dc.date.updated | 2024-04-01T07:53:38Z | en |
| dc.description.abstract | Sentiment Analysis (SA) is an essential task for numerous realworld applications. However, the majority of SA research focuses on high-resource languages such as English and Chinese, while limited-resource languages like Arabic and Bengali receive less attention. Additionally, existing Arabic sentiment analysis methods based on advanced artificial intelligence (AI) approaches tend to operate as black boxes, making it challenging to comprehend the reasoning behind their predictions. This paper proposes an explainable sentiment classification framework for the Arabic language. We introduce a noise layer to different deep learning (DL) models, including BiLSTM and CNN-BiLSTM, to address the issue of overfitting. The proposed framework enables the explanation of specific predictions by training a local surrogate explainable model, shedding light on the reasons behind each sentiment prediction (positive or negative). Experiments were conducted on publicly available benchmark Arabic SA datasets, and the results demonstrated that the inclusion of noise layers in the DL model improves performance for the Arabic language by mitigating overfitting. Our method also outperformed several state-of-the-art approaches. Moreover, the introduction of explainability with the noise layer enhances transparency and accountability, making the model suitable for practical adoption in AI-enabled systems. | en |
| dc.description.version | Published version | en |
| dc.format.mimetype | application/pdf | en |
| dc.identifier.doi | https://doi.org/10.1145/3639233.3639241 | en |
| dc.identifier.uri | https://hdl.handle.net/10919/118485 | en |
| dc.language.iso | en | en |
| dc.publisher | ACM | en |
| dc.rights | Creative Commons Attribution 4.0 International | en |
| dc.rights.holder | The author(s) | en |
| dc.rights.uri | http://creativecommons.org/licenses/by/4.0/ | en |
| dc.title | Arabic Sentiment Analysis with Noisy Deep Explainable Model | en |
| dc.type | Article - Refereed | en |
| dc.type.dcmitype | Text | en |