Automated Classification and Sentiment Analysis of Feedback on Education Policies using Text Mining Techniques

• Chunming Chen,
• Wei Zhu,

Vol. 19, No. 11, November 30, 2025
10.3837/tiis.2025.11.011, Download Paper (Free):

• education policy
• Dual-layer long short-term memory networks
• Natural Language Processing
• sentiment analysis
• Text Mining

Abstract

In modern education, student feedback is a critical resource for shaping policies. However, analyzing unstructured feedback from multiple institutions is time-consuming, labor-intensive, and prone to human bias. To address this, we propose the Dual-Layer Long Short-Term Memory (D2LSTM) model for automated sentiment analysis of educational policy feedback. The process begins with Natural Language Processing (NLP) techniques, including text preprocessing, tokenization, and Word2Vec embeddings. Feature extraction is performed using TF-IDF and Bag-of-Words to capture both word frequency and contextual meaning. The D2LSTM architecture is designed to model long-term dependencies and contextual nuances, enabling accurate classification of feedback into positive, neutral, or negative sentiments. The model was trained on feedback data from prominent universities in North India, covering six categories of educational policy. Implementation used Python and the Natural Language Toolkit (NLTK). Performance was evaluated against existing models across sentiment and category-based analyses using accuracy, precision, recall, and F1-score. Results show that D2LSTM outperforms traditional methods, offering a scalable and bias-reduced approach. By providing actionable insights from student feedback, the model supports data-driven policy decisions, demonstrating strong potential for real-world deployment in educational settings.

Statistics

Cite this article