Combining Active Learning and Task Adaptation with BERT for Cost-Effective Annotation of Social Media Datasets.
Jens LemmensWalter DaelemansPublished in: WASSA@ACL (2023)
Keyphrases
- cost effective
- active learning
- social media
- machine learning
- low cost
- annotation effort
- social media data
- cost effectiveness
- semi supervised
- training examples
- social networks
- semi supervised learning
- automatic annotation
- cost sensitive
- data center
- learning process
- learning strategies
- labeled data
- training data
- supervised learning
- batch mode
- training set
- big data
- data sets
- random sampling
- learning algorithm
- metadata
- user generated content
- class imbalance
- semantic annotation
- data management
- relevance feedback
- long term