Summary of the Paper:
"Sentiment Analysis: A Survey on Design Framework, Applications, and Future
Scopes"
Authors:
Monali Bordoloi and Saroj Kumar Biswas
Published:
Artificial Intelligence Review, 2023
Key Contributions:
1. Comprehensive Framework: Proposes a systematic, step-by-step framework
for designing sentiment analysis models, covering data collection, pre-
processing, feature extraction, sentiment classification, and summarization.
2. Granularity Levels: Discusses sentiment analysis at multiple levels (document,
sentence, word, aspect, concept, user, clause, and sense levels) with examples.
3. Techniques and Algorithms: Reviews machine learning (SVM, Naive Bayes,
deep learning), lexicon-based (dictionary and corpus approaches), and hybrid
methods for sentiment classification.
4. Applications: Highlights diverse applications, including product reviews,
political tweets, crisis management, stock market prediction, and movie
reviews.
5. Performance Metrics: Evaluates models using accuracy, precision, recall, F-
measure, MCC, and AUC-ROC.
6. Future Directions: Identifies gaps like domain dependency, sarcasm detection,
and advanced pre-processing techniques, suggesting improvements for future
research.
Structure:
• Introduction: Defines sentiment analysis and its importance in decision-
making.
• Background: Explores historical context and relationships with NLP,
computational linguistics, and text mining.
• Framework: Outlines modules from data collection to sentiment
summarization.
• Classification Techniques: Compares supervised (SVM, Naive Bayes, deep
learning) and unsupervised methods.
• Applications: Categorizes use cases by data content (e.g., social media,
reviews).
• Challenges: Addresses issues like negation handling and cross-domain
adaptability.
• Conclusion: Emphasizes the need for robust models and interdisciplinary
applications.
Significance:
This survey serves as a guide for researchers and practitioners, offering insights into
methodologies, tools, and emerging trends in sentiment analysis. It bridges gaps in
existing literature by providing a holistic view of the field and suggesting future
research directions.
Keywords: Sentiment Analysis, NLP, Machine Learning, Lexicon-Based Approaches,
Deep Learning, Applications.
