Table 1 Research work related to machine learning classifiers for sentiment analysis
Author/year | Technical approach | Accuracy in % | Dataset domain |
|---|---|---|---|
Pang et al. (2002) [25] | Applied N-gram model with NB, SVM, ME | 77.4–82.9 | Internet Movie Database (IMDb) |
Dave et al. (2003) [8] | Used N-gram model for feature extraction with SVM, NB classifier | 87.0 | Product review from Amazon and CNET |
Annett and Kondrak (2008) [9] | Considered WordNet as Lexical resource with SVM, NB, Decision Tree classifier | 75.0 | Movie reviews (IMDb)- 1000 (+) and 1000 (−) reviews |
Ye et al. (2009) [12] | NB, SVM classifier used for classification | 85.14 | Travel blogs |
Mouthami et al. (2013) [11] | TF-IDF and POS tagging with fuzzy classification algorithm | 87.4 | Movie review dataset |
Zha et al. (2014) [17] | SVM, NB, ME classifier adopted with evaluation matrices F1-Measure | 83.0–88.43 | Customer reviews (feedback) |
Habernal et al. (2014) [26] | N-gram and POS related features and emoticons are selected using MI, CHI, OR, RS method. Classifier ME and SVM used for classification | 78.50 | Dataset from social media |
Zhang et al. (2015) [10] | Use word2vec for features with SVM classifier for classification | 89.95–90.30 | Chinese review dataset |
Luo et al. (2016) [21] | First transform the text into low dimensional emotional space (ESM), next implement SVM, NB, DT classifier | 63.28–79.21 | Stock message text data |