LETS: A Label-Efficient Training Scheme for Aspect-Based Sentiment Analysis by Using a Pre-Trained Language Model.
Heereen ShimDietwig LowetStijn LucaBart VanrumstePublished in: IEEE Access (2021)
Keyphrases
- language model
- sentiment analysis
- pre trained
- language modeling
- n gram
- opinion mining
- text classification
- sentiment classification
- speech recognition
- probabilistic model
- document retrieval
- sentiment lexicon
- training examples
- retrieval model
- test collection
- sentence level
- information retrieval
- query expansion
- natural language processing
- product reviews
- text mining
- training set
- training data
- relevance model
- translation model
- product features
- statistical machine translation
- feature selection