A novel text sentiment analysis system using improved depthwise separable convolution neural networks.

Area of Science:

• Computational Linguistics
• Artificial Intelligence
• Natural Language Processing

Background:

• Human emotions significantly influence behavior, making emotion classification crucial for prediction and decision-making.
• The exponential growth of online text data overwhelms manual classification, necessitating efficient computational methods for sentiment analysis.
• Existing deep learning models for sentiment analysis suffer from high complexity and inadequate utilization of linguistic features, including word vectors.

Purpose of the Study:

• To address the limitations of current deep learning models in text-based sentiment analysis.
• To develop a more efficient and accurate method for extracting emotional tendencies from text data.
• To improve the extraction of word vector and context information while reducing model complexity.

Main Methods:

• An upgraded convolutional neural network (CNN) model was employed.
• Text separable convolution algorithm was utilized for hierarchical convolution on text features.
• The proposed model integrates refined extraction of word vector and context information, avoiding semantic confusion and reducing network complexity.

Main Results:

• The improved CNN model demonstrated superior performance in text-based sentiment analysis tasks compared to other models.
• The application of text separable convolution enhanced the extraction of linguistic features.
• The model effectively reduced complexity and semantic confusion in sentiment analysis.

Conclusions:

• The proposed upgraded CNN model offers a valuable advancement for text-based sentiment analysis.
• The method provides a more efficient and accurate approach to understanding emotional tendencies in large text datasets.
• This research contributes significantly to the field of sentiment analysis and its applications.