Scholarlink

<p>Normally, people’s opinions and thoughts are evaluated for sentiment analysis. Based on the sentiment analysis process, the customer feelings are analyzed by the organization with the wealth of online data. In addition to this, it is significant to perform an emotional analysis of the written reviews to improve customer satisfaction with the restaurant’s services. The text segmentation is performed by various researchers based on the availability of different computerized methods and massive data. People’s opinions are explored mainly by the sentiment classification task to make better decisions in the future. The user experience is expressed mainly in social media in both text and image formats. These results indicate difficulty in understanding the variety of activities based on sentiment classification. Machine learning models used in sentiment classification tend to increase time complexity; therefore, it is significant to tackle various issues in the existing research framework. Therefore, this study aims to design a novel multimodal food review framework with deep learning techniques. Initially, essential multimodal data, like images and texts for validation, are collected from online platforms. Subsequently, the collected data are provided for the text’s pre-processing phase. Then, the pre-processed texts are utilized in the feature extraction phase, and also the significant features presented in the pre-processed texts are extracted using the Bidirectional Encoder Representations from Transformers (BERT) technique, and these features are considered as feature set 1. Further, from the collected images, a second set of features is extracted through the Vision Transformer (ViT) mechanism. Later, the two sets of extracted features are given to the mutual-information-gain-based weighted-fused-feature-selection phase. In this phase, weights are added with the selected fused features, and the weights are optimized using the Iteration Condition of Secretary Bird Optimization (ICSBO). Here, the developed Adaptive Deep Bidirectional Residual Recurrent Neural Network (ADBi-R2NN) technique is employed to make better decisions about food reviews. Furthermore, some parameters in ADBi-R2NN are optimized using an ICSBO to achieve more accurate food review outcomes. In the end, the experiments are conducted to evaluate the sentiment classification model.</p>

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

Development of Adaptive Deep Bidirectional Residual RNN for Sentiment Analysis in Food Review

  • 期刊论文

  • 发表日期：

    2026年5月7日

查看全文
  • Akkireddi Vara Prasad,
  • K. Vedavathi

摘要

Normally, people’s opinions and thoughts are evaluated for sentiment analysis. Based on the sentiment analysis process, the customer feelings are analyzed by the organization with the wealth of online data. In addition to this, it is significant to perform an emotional analysis of the written reviews to improve customer satisfaction with the restaurant’s services. The text segmentation is performed by various researchers based on the availability of different computerized methods and massive data. People’s opinions are explored mainly by the sentiment classification task to make better decisions in the future. The user experience is expressed mainly in social media in both text and image formats. These results indicate difficulty in understanding the variety of activities based on sentiment classification. Machine learning models used in sentiment classification tend to increase time complexity; therefore, it is significant to tackle various issues in the existing research framework. Therefore, this study aims to design a novel multimodal food review framework with deep learning techniques. Initially, essential multimodal data, like images and texts for validation, are collected from online platforms. Subsequently, the collected data are provided for the text’s pre-processing phase. Then, the pre-processed texts are utilized in the feature extraction phase, and also the significant features presented in the pre-processed texts are extracted using the Bidirectional Encoder Representations from Transformers (BERT) technique, and these features are considered as feature set 1. Further, from the collected images, a second set of features is extracted through the Vision Transformer (ViT) mechanism. Later, the two sets of extracted features are given to the mutual-information-gain-based weighted-fused-feature-selection phase. In this phase, weights are added with the selected fused features, and the weights are optimized using the Iteration Condition of Secretary Bird Optimization (ICSBO). Here, the developed Adaptive Deep Bidirectional Residual Recurrent Neural Network (ADBi-R2NN) technique is employed to make better decisions about food reviews. Furthermore, some parameters in ADBi-R2NN are optimized using an ICSBO to achieve more accurate food review outcomes. In the end, the experiments are conducted to evaluate the sentiment classification model.