Covers: theory of Shrinking the Search Space for Seq2Seq Models
Estimated time needed to finish: 45 minutes
Questions this item addresses:
  • What is the theory behind shrinking the search space for seq2seq models?
How to use this item?

Read pages 1-8 of the paper.

Author(s) / creator(s) / reference(s)
Ben Peters and Andre F. T. Martins
0 comment
Recipe
publicShare
Star(0)

Understanding the Paper, "Smoothing and Shrinking the Sparse Seq2Seq Search Space," Part 2

Contributors
Total time needed: ~2 hours
Objectives
Help the reader understand the paper in question, as well as recent trends in seq2seq modeling, decoders, and loss functions
Potential Use Cases
Needing to improve a seq2seq model for downstream tasks like translation, inflection, and pronunciation
Who is This For ?
INTERMEDIATENLP practitioners working on seq2seq models
Click on each of the following annotated items to see details.
OTHER 1. What happened in Natural language generation decoders in 2019?
  • What is the context for the paper?
  • What developments have there been in decoders in NLP in 2019?
5 minutes
ARTICLE 2. What is Label Smoothing?
  • What is the label smoothing discussed in the paper that is the subject of this RECIPE?
20 minutes
PAPER 3. Learning with Fenchel-Young Losses
  • What are the Fenchel-Young Losses mentioned in the paper that is the subject of this RECIPE?
5 minutes
PAPER 4. On NMT Search Errors and Model Errors: Cat Got Your Tongue?
  • What is the cat got your tongue problem in neural machine translation?
5 minutes
PAPER 5. Sparse Sequence-to-Sequence Models
  • What is the background to the newer paper by these authors that is the subject of this RECIPE?
  • What are sparse and dense sequence to sequence models?
10 minutes
PAPER 6. Smoothing and Shrinking the Sparse Seq2Seq Search Space
  • What is the theory behind shrinking the search space for seq2seq models?
45 minutes
REPO 7. Smoothing and Shrinking the Sparse Seq2Seq Search Space
  • How was label smoothing implemented in the paper that is the subject of this RECIPE?
20 minutes
OTHER 8. Morphological Inflection
  • What is morphological inflection?
1 minutes
PAPER 9. Grapheme-to-Phoneme Conversion with Convolutional Neural Networks
  • What is grapheme-to-phoneme (G2P) conversion?
1 minutes

Concepts Covered

0 comment