People Make Better Edits: Measuring the Efficacy of LLM-Generated Counterfactually Augmented Data for Harmful Language Detection.
Indira SenDennis AssenmacherMattia SamoryIsabelle AugensteinWil van der AalstClaudia WagnerPublished in: CoRR (2023)
Keyphrases
- synthetic data
- data collection
- data processing
- original data
- data points
- data sets
- raw data
- statistical analysis
- computer systems
- false alarms
- database
- data quality
- image data
- data analysis
- small number
- programming language
- input data
- probability distribution
- data sources
- data structure
- training data
- detection rate
- knowledge base
- social networks
- information retrieval
- neural network