The Scalable Fuzzy Inference-Based Ensemble Method for Sentiment Analysis

Area of Science:

• Natural Language Processing
• Machine Learning
• Data Science

Background:

• The internet generates vast amounts of user-generated text data (social media, blogs, news).
• Institutions need efficient methods to analyze this big data for public opinion and sentiment.
• Existing sentiment analysis methods can be improved for accuracy and scalability.

Purpose of the Study:

• To propose a scalable sentiment classification framework using a fuzzy inference mechanism.
• To enhance sentiment analysis by ensembling diverse methods like dictionary-based, word embedding, and count vectorization.
• To improve upon classical ensemble methods by incorporating weighted base learners and fuzzy rules.

Main Methods:

• A fuzzy inference system was designed to evaluate compound probability scores from sentiment analysis techniques.
• The system integrates valence-aware dictionaries, word embeddings, and count vectorization.
• A novel ensemble approach allows weighting of base learners and combines algorithms using fuzzy rules.

Main Results:

• The proposed framework was tested on four tagged social network datasets.
• Experimental results demonstrated improved accuracy for both 2-class (positive/negative) and 3-class (positive/neutral/negative) sentiment classification.
• The fuzzy inference mechanism effectively combined strengths of different sentiment analysis algorithms.

Conclusions:

• The developed scalable framework offers improved accuracy for sentiment classification.
• The fuzzy ensemble approach provides a robust method for analyzing large volumes of text data.
• This research contributes to more effective big data analysis for understanding public opinion.