Scholarlink

The expansion of social media platforms has given users a powerful platform to express their opinions on a wide range of topics, significantly influencing the field of sentiment analysis (SA) within natural language processing (NLP). While SA is essential for deriving insights and informing decision-making based on public sentiment, it encounters specific challenges with the Arabic language due to its diverse dialects and complex morphology. Deep learning (DL) techniques, particularly convolutional neural networks (CNNs) and long short-term memory (LSTM) networks, have greatly improved SA by capturing critical features more effectively than traditional machine learning (ML) approaches. This paper explores two key architectures: the CNN approach, which utilizes the FastText word embedding model, and the LSTM approach, also incorporating FastText, for Arabic sentiment analysis (ASA). The study assesses the performance of these models across two distinct datasets and validates the findings of prior comparative research. The results show that both models perform robustly, with higher classification accuracy observed on the first dataset compared to the second, reinforcing the conclusions drawn from earlier research in aspect-based sentiment analysis.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

In-Depth Evaluation of Leading Neural Network Models with Word Embedding Approach for Arabic Sentiment Analysis

  • 章节

  • 发表日期：

    2025年

查看全文
  • Youssra Zahidi,
  • Yassine Al-Amrani

摘要

The expansion of social media platforms has given users a powerful platform to express their opinions on a wide range of topics, significantly influencing the field of sentiment analysis (SA) within natural language processing (NLP). While SA is essential for deriving insights and informing decision-making based on public sentiment, it encounters specific challenges with the Arabic language due to its diverse dialects and complex morphology. Deep learning (DL) techniques, particularly convolutional neural networks (CNNs) and long short-term memory (LSTM) networks, have greatly improved SA by capturing critical features more effectively than traditional machine learning (ML) approaches. This paper explores two key architectures: the CNN approach, which utilizes the FastText word embedding model, and the LSTM approach, also incorporating FastText, for Arabic sentiment analysis (ASA). The study assesses the performance of these models across two distinct datasets and validates the findings of prior comparative research. The results show that both models perform robustly, with higher classification accuracy observed on the first dataset compared to the second, reinforcing the conclusions drawn from earlier research in aspect-based sentiment analysis.