Low Frequency Names Exhibit Bias and Overfitting in Contextualizing Language Models.
Robert WolfeAylin CaliskanPublished in: EMNLP (1) (2021)
Keyphrases
- language model
- low frequency
- high frequency
- language modeling
- frequency domain
- wavelet transform
- n gram
- speech recognition
- document retrieval
- probabilistic model
- language modelling
- retrieval model
- subband
- statistical language models
- query expansion
- wavelet coefficients
- test collection
- discrete wavelet transform
- decision trees
- information retrieval
- language models for information retrieval
- computer vision
- pseudo relevance feedback
- fusion rules
- context sensitive
- query terms
- bayesian networks
- smoothing methods
- high frequency components
- high resolution
- information fusion
- named entities
- vector space model
- image data
- multiresolution
- feature vectors
- okapi bm
- low and high frequency