Beyond 512 Tokens: Siamese Multi-depth Transformer-based Hierarchical Encoder for Long-Form Document Matching.
Liu YangMingyang ZhangCheng LiMichael BenderskyMarc NajorkPublished in: CIKM (2020)
Keyphrases
- matching algorithm
- information retrieval systems
- fuzzy logic
- bit rate
- web documents
- motion estimation
- document classification
- information retrieval
- rate distortion
- document retrieval
- pattern matching
- hierarchically organized
- concept hierarchy
- document representation
- depth information
- text documents
- relevant documents
- document collections
- co occurrence
- keywords
- high quality
- computer vision