Feature Selection for Highly Skewed Sentiment Analysis Tasks.
Can LiuSandra KüblerNing YuPublished in: SocialNLP@COLING (2014)
Keyphrases
- sentiment analysis
- text classification
- highly skewed
- feature selection
- opinion mining
- sentiment classification
- naive bayes
- text categorization
- multi task
- sentence level
- product reviews
- text mining
- multi class
- imbalanced data
- support vector
- machine learning
- class distribution
- natural language processing
- transfer learning
- mutual information
- knn
- feature space
- sentiment lexicon
- class imbalance
- classification accuracy
- neural network