A Sentiment Analysis Method Based on a Blockchain-Supported Long Short-Term Memory Deep Network

Area of Science:

• Artificial Intelligence
• Computer Science
• Data Science

Background:

• Traditional sentiment analysis methods face challenges due to technical processing diversity and cultural influences, leading to variable results.
• Existing machine learning and deep learning models often lack a robust framework for data integrity and cross-cultural applicability.

Purpose of the Study:

• To develop an enhanced sentiment analysis framework integrating blockchain technology with Long Short-Term Memory (LSTM) architecture.
• To address cultural temperament effects and improve the accuracy and reliability of sentiment analysis.

Main Methods:

• Integration of a blockchain layer with an LSTM architecture to create a secure and verifiable machine learning process.
• Utilizing a 'Proof of Learning' consensus mechanism for data management and integrity within the machine learning workflow.
• Application of the proposed model to a Twitter dataset analyzing negative, neutral, and positive sentiments.

Main Results:

• The proposed blockchain-integrated LSTM model achieved a significantly higher accuracy rate of 92.85% on the Twitter dataset.
• This represents a substantial improvement over previous methods, which reported accuracies of 14% and 63% in different cultural contexts.
• The framework ensures cryptographic connection and metadata transfer, enhancing the reliability of the learning database.

Conclusions:

• The integration of blockchain with LSTM offers a robust solution for accurate and culturally unbiased sentiment analysis.
• The 'Proof of Learning' framework provides a verifiable mechanism for machine learning processes and data management.
• This approach demonstrates a promising advancement in overcoming the limitations of traditional sentiment analysis techniques.