Advanced Natural Language Processing (NLP) Techniques for Text-Data Based Sentiment Analysis on Social Media

Journal of Artificial Intelligence and Big Data | Vol 1, Issue 1

Table 1. Summary of Sentiment ClassificationTechniques in Social Media Using Machine Learning

Paper	Method	Dataset	Key Findings	Limitations &Future Work
Kanakaraj and Guddeti (2015)	NLP techniques, Word Sense Disambiguation, Ensemble classification	Twitter posts on news events	Ensemble classification improves accuracy by 3-5% over traditional ML classifiers	Future work could explore deep learning models for further accuracy enhancement
Chirawichitchai (2014)	Term weighting, SVM, Information Gain feature selection	Thai text dataset	Boolean weighting with SVM achieves the highest accuracy (77.86%)	Future work can focus on expanding emotion classification for multilingual settings
Hogenboom et al. (2014)	Spreading sentiment lexicon and cross-linguistic sentiment mapping	English and Dutch language datasets	Sentiment propagation improves accuracy by up to 47%	Further research can investigate additional languages and domain-specific sentiment lexicons
Anjaria and Guddeti (2014)	Supervised ML (SVM, Naïve Bayes, ANN), Unigram & Bigram features, Influence Factor	Twitter statistics (Karnataka State Assembly Elections 2013, US Presidential Elections 2012)	SVM achieved highest accuracy (88% for US Elections, 68% for Indian Elections)	Future work can incorporate deep learning models and social influence factors for better prediction
Volkova, Wilson, and Yarowsky (2013)	Understanding how gender differs in the classification of sentiment, polarity, and subjectivity	English, Spanish, and Russian Twitter data	Gender-based language differences improve polarity classification (2.5-5% improvement in F-measure)	Future studies can explore additional cultural and linguistic variations for sentiment analysis