MockingBERT: A Method for Retroactively Adding Resilience to NLP Models.
Jan JezabekAkash SinghPublished in: COLING (2022)
Keyphrases
- bayesian framework
- probabilistic model
- preprocessing
- machine learning methods
- statistical model
- pairwise
- cost function
- similarity measure
- modeling method
- detection method
- high accuracy
- objective function
- prior knowledge
- experimental evaluation
- multiscale
- statistical methods
- high precision
- classification method
- missing data
- parameter estimation
- support vector machine
- significant improvement
- dynamic programming
- segmentation algorithm
- artificial neural networks
- least squares
- mathematical model
- natural language
- knowledge representation
- computational cost
- prediction model
- monte carlo simulation
- feature extraction