Evaluation of adjective and adverb types for effective Twitter sentiment classification

Area of Science:

• Natural Language Processing
• Machine Learning
• Computational Linguistics

Background:

• Twitter is a major platform for user sentiment expression.
• Sentiment analysis classifies tweets to understand user opinions.
• Feature selection, particularly parts of speech (POS), is crucial for sentiment classification.

Purpose of the Study:

• To investigate the impact of various adjective and adverb types on sentiment classification accuracy.
• To analyze the effectiveness of combining different adjective and adverb types.
• To compare the proposed method against benchmark datasets and existing models.

Main Methods:

• Utilized a human-annotated tweet dataset for sentiment analysis.
• Examined specific types of adjectives (e.g., superlative) and adverbs (e.g., comparative).
• Evaluated combinations of adjective and adverb types for improved classification.

Main Results:

• Achieved up to 83% classification accuracy using general superlative adjectives.
• Demonstrated significant accuracy improvements with comparative general adverbs.
• Combinations of general adjectives and adverbs also substantially improved results.

Conclusions:

• Unexplored adjective and adverb types offer superior accuracy compared to state-of-the-art probabilistic models.
• The proposed model reduces reliance on lexicon-based dictionaries, speeding up analysis.
• This research provides valuable insights for sentiment analysis and efficient data processing.