The Low-Dimensional Linear Geometry of Contextualized Word Representations.
Evan HernandezJacob AndreasPublished in: CoNLL (2021)
Keyphrases
- low dimensional
- linear subspace
- appearance manifolds
- high dimensional
- dimensionality reduction
- high dimensional data
- manifold learning
- dimension reduction
- underlying manifold
- co occurrence
- vector space
- euclidean space
- principal component analysis
- input space
- three dimensional
- n gram
- geometric structure
- feature space
- word sense disambiguation
- multidimensional scaling
- word recognition
- word clouds
- data points
- feature representation
- nonlinear dimensionality reduction
- graph embedding
- video sequences
- information retrieval
- high dimensional image space