ViBERTgrid: A Jointly Trained Multi-Modal 2D Document Representation for Key Information Extraction from Documents.
Weihong LinQifang GaoLei SunZhuoyao ZhongKai HuQin RenQiang HuoPublished in: CoRR (2021)
Keyphrases
- multi modal
- document representation
- information extraction
- text documents
- web documents
- document clustering
- document collections
- bag of words
- text mining
- vector space model
- document content
- index terms
- document categorization
- language model
- vector space
- data fusion
- semantic information
- information retrieval
- text data
- text classification
- machine learning
- precision and recall
- high dimensional
- natural language processing
- named entities
- semantic similarity
- topic models
- background knowledge
- keywords
- document retrieval
- feature selection
- computer vision
- image processing
- feature extraction
- image representation
- training set
- data mining