Research Paper

Note: Store the paper I read already. Put to Pending material if not read yet.

A repo + Kaggle notebook explaining Key Concepts of various papers, e.g. BERT, Donut…etc

NLP

Chargrid: Towards Understanding 2D Documents(2018)
- Chargrid representation :
  - All characters are 1-hot encoded into a mapping integer
  - i.e. the encoding is NOT contextualized
- https://github.com/sciencefictionlab/chargrid-pytorch
BERTgrid: Contextualized Embedding for 2D Document Representation and Understanding(2019)
- BERTgrid representation :
  - similar to Chargrid, but it is word level & all words are embedded (an array of values) instead of only representing the char as an integer.
  - i.e. the embedding is contextualized