Scholarlink

With the fast distribution of the social networks, more users begin to freely express their opinions on different media such as Twitter and Facebook. Emotions, opinions, and information from daily lives are shared by people on these social media network. This study investigates the performance of deep residual network (ResNet) on Arabic sentiment analysis using Word2Vec for text representation. Two datasets are used for the evaluation of the model; ASTD standard dataset consists of 10,000 Arabic Tweets, and a new artificial dataset collected from Twitter in Jordan, it includes 7100 Tweets related to different topics annotated after preprocessing with four sentiments: positive, negative, objective, and neutral. Experimental results proved that the ResNet model obtained an accuracy of 96.59% and 80.30% on the two datasets, respectively. Furthermore, the results showed a preference for the proposed method results comparing to previous works.

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

Sentiment Analysis of Arabic Tweets Using Deep Residual Neural Networks

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Ahmad T. Al-Taani,
  • Razan M. Al-Dheirat

摘要

With the fast distribution of the social networks, more users begin to freely express their opinions on different media such as Twitter and Facebook. Emotions, opinions, and information from daily lives are shared by people on these social media network. This study investigates the performance of deep residual network (ResNet) on Arabic sentiment analysis using Word2Vec for text representation. Two datasets are used for the evaluation of the model; ASTD standard dataset consists of 10,000 Arabic Tweets, and a new artificial dataset collected from Twitter in Jordan, it includes 7100 Tweets related to different topics annotated after preprocessing with four sentiments: positive, negative, objective, and neutral. Experimental results proved that the ResNet model obtained an accuracy of 96.59% and 80.30% on the two datasets, respectively. Furthermore, the results showed a preference for the proposed method results comparing to previous works.