Abstract
Sentiment classification, which is usually employed to categorize emotions in social media text, has been extensively studied for public opinion monitoring. For example, it can be used to examine feedback from customers on online shopping platforms to assess their satisfaction levels. Sentiment classification achieves high accuracy with the English language. In contrast, sentiment classification in Arabic falls behind, mainly due to the intricate nature of Arabic text, which poses difficulties for contemporary machine learning methods. This study evaluates the effectiveness of the Pathways Language Model (PaLM) and CAMeLBERT to show efficiency of large language models in detecting sentiment in Arabic social media. The results indicate that a pre-trained model using the BERT (CAMeLBERT) architecture outperformed PaLM, achieving an accuracy of 63.87% compared to 61.01% with PaLM. These results underscore the ongoing difficulties that modern language models encounter in accurately interpreting the subtleties of Arabic sentiment classification.
Recommended Citation
Almutrash, Salman and Abudalfa, Shadi, "Comparative Study on the Efficiency of Using PaLM and
CAMeLBERT for Arabic Entity Sentiment Classification" (2024). SaudiCIS 2024 Proceedings. 66.
https://aisel.aisnet.org/saudicis2024/66