Lacuna Reconstruction: Self-Supervised Pre-Training for Low-Resource Historical Document Transcription.
Nikolai VoglerJonathan Parkes AllenMatthew Thomas MillerTaylor Berg-KirkpatrickPublished in: NAACL-HLT (Findings) (2022)
Keyphrases
- training process
- image reconstruction
- training set
- high resolution
- information retrieval
- document classification
- retrieval systems
- information retrieval systems
- reconstruction process
- database
- training examples
- web documents
- supervised learning
- keywords
- document clustering
- vector space model
- resource management
- document retrieval
- resource allocation
- test set
- document collections
- online learning
- three dimensional
- document images
- learning algorithm
- web resources
- document representation
- texture synthesis
- structured documents
- reconstruction method
- text classifiers
- artificial neural networks