Scholarlink

<p>Opinion Mining is a relatively recent area of research for harnessing user data from heterogeneous online resources to comprehend public sentiment better, make well-informed decisions, and adjust their strategy in response to real-time feedback. Therefore, accurate sentiment analysis plays a vital role. The advantages of this study are two-fold; it will not only help the teaching and administrative staff in determining the current sentimental condition of the students, but the teaching staff would also be able to improve their course contents given students’ feedback. The proposed methodology implements the Multi-Layer Perceptron (MLP), Gated Recurrent Units (GRU), and Long Short Term Memory (LSTM) on dual datasets. The proposed model is tested on the locally developed dataset “DiskusEdu” based on education feedback tweets and a general dataset named “Twitter Sentiment”. The proposed methodology presented a promising result, representing better performance than existing techniques. The accuracies of MLP on the “DiskusEdu” dataset with Unigram and Bigram Associations are 0.74 and 0.83, respectively. The accuracies of the GRU and LSTM on the “Twitter Sentiment” dataset are 0.89 and 0.88, respectively. In modern interactive learning systems, teachers often want to know the sentimental condition of their students before starting their lectures. This research works as a source of feedback for teaching staff and the administration staff at the institution to streamline academic activities like teacher performance, lecture contents and schedule, and many others.</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

Leveraging deep learning for sentiment analysis: insights from academic and general feedback across dual datasets

  • 期刊论文

  • 发表日期：

    2025年8月9日

查看全文
  • Muhammad Munwar Iqbal,
  • Waqas Mehmood,
  • Shabana Ramzan,
  • Muhammad Faheem Mushtaq

摘要

Opinion Mining is a relatively recent area of research for harnessing user data from heterogeneous online resources to comprehend public sentiment better, make well-informed decisions, and adjust their strategy in response to real-time feedback. Therefore, accurate sentiment analysis plays a vital role. The advantages of this study are two-fold; it will not only help the teaching and administrative staff in determining the current sentimental condition of the students, but the teaching staff would also be able to improve their course contents given students’ feedback. The proposed methodology implements the Multi-Layer Perceptron (MLP), Gated Recurrent Units (GRU), and Long Short Term Memory (LSTM) on dual datasets. The proposed model is tested on the locally developed dataset “DiskusEdu” based on education feedback tweets and a general dataset named “Twitter Sentiment”. The proposed methodology presented a promising result, representing better performance than existing techniques. The accuracies of MLP on the “DiskusEdu” dataset with Unigram and Bigram Associations are 0.74 and 0.83, respectively. The accuracies of the GRU and LSTM on the “Twitter Sentiment” dataset are 0.89 and 0.88, respectively. In modern interactive learning systems, teachers often want to know the sentimental condition of their students before starting their lectures. This research works as a source of feedback for teaching staff and the administration staff at the institution to streamline academic activities like teacher performance, lecture contents and schedule, and many others.