Modified term frequency-inverse document frequency based deep hybrid framework for sentiment analysis.

Area of Science:

• Natural Language Processing
• Artificial Intelligence
• Machine Learning

Background:

• Sentiment analysis is vital for understanding public opinion from user-generated text.
• Existing methods often require improvements in feature extraction and representation for better accuracy.

Purpose of the Study:

• To propose a novel hybridized neural network framework for enhanced sentiment analysis.
• To improve text feature vectorization using a modified Term Frequency-Inverse Document Frequency (TF-IDF) approach.

Main Methods:

• Preprocessing text data and applying a modified TF-IDF scheme with a non-linear global weighting factor.
• Utilizing k-best selection for text feature vectorization and pre-trained embeddings for mathematical representation.
• Employing a deep neural network combining Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) for sentiment classification.

Main Results:

• The proposed model demonstrated superior performance compared to state-of-the-art baseline models.
• The hybridized approach effectively captured local and historical features for accurate sentiment polarization.
• Validation across multiple datasets confirmed the model's efficacy and robustness.

Conclusions:

• The developed sentiment analysis framework offers a significant advancement in extracting public sentiment.
• The integration of modified TF-IDF, embeddings, and a CNN-LSTM network provides a powerful tool for opinion mining.
• This research contributes to more effective and accurate sentiment analysis in Natural Language Processing.