Does Noise Really Matter? Investigation into the Influence of Noisy Labels on BERT-Based Question Answering System.
Dmitriy AlexandrovAnastasiia ZakharovaNikolay ButakovPublished in: Int. J. Semantic Comput. (2024)
Keyphrases
- noisy data
- noisy environments
- noise free
- label noise
- measurement noise
- high noise
- low snr
- missing data
- random noise
- noise reduction
- question answering
- noise level
- image noise
- low signal to noise ratio
- noisy observations
- additive noise
- noise model
- median filter
- multi label
- speech signal
- active learning
- pairwise
- salt and pepper noise
- training data
- arbitrary shape
- clustering algorithm
- factors influencing
- image structure
- noisy images
- degraded images
- signal to noise ratio
- class noise
- image segmentation