The Use of Unlabeled Data Versus Labeled Data for Stopping Active Learning for Text Classification.
Garrett BeattyEthan KochisMichael BloodgoodPublished in: ICSC (2019)
Keyphrases
- labeled data
- text classification
- unlabeled data
- active learning
- semi supervised learning
- transfer learning
- labeled and unlabeled data
- unlabeled examples
- labeled examples
- semi supervised classification
- machine learning
- labeled training data
- text categorization
- training examples
- co training
- unlabeled instances
- semi supervised
- naive bayes
- number of labeled examples
- feature selection
- text mining
- domain adaptation
- unlabeled documents
- positive examples
- class labels
- knn
- label propagation
- labeled instances
- unsupervised learning
- labeling effort
- text classifiers
- abundant unlabeled data
- document classification
- pairwise constraints
- sentiment classification
- fully supervised
- supervised learning algorithms
- learning algorithm
- training set
- semi supervised learning methods
- selective sampling
- supervised methods
- learning tasks
- cost sensitive
- target domain
- multi label
- multi view
- semi supervised learning algorithms
- supervised learning
- data points
- semisupervised learning
- reinforcement learning
- query by committee
- label information