Domain Adaptation of Transformer-Based Models using Unlabeled Data for Relevance and Polarity Classification of German Customer Feedback.
Ahmad Idrissi-YaghirHenning SchäferNadja BauerChristoph M. FriedrichPublished in: CoRR (2022)
Keyphrases
- domain adaptation
- unlabeled data
- labeled data
- semi supervised learning
- semi supervised
- co training
- target domain
- transfer learning
- active learning
- cross domain
- training data
- text classification
- sentiment classification
- polarity classification
- multiple sources
- supervised learning
- prior knowledge
- model selection
- test data
- document classification
- text categorization
- labeled and unlabeled data
- source domain
- labeled training data
- data points
- training set
- class labels
- probabilistic model
- information retrieval
- data mining
- unsupervised learning
- labeled examples
- source and target domains