1 of 7

NLU Lab: Paper Reading (14 Feb 2024)

Today:

  • What is a (NLP) paper?
  • Why are they written & read
  • How to read them (effectively)

Right for the Wrong Reasons: Diagnosing Syntactic Heuristics in Natural Language Inference

(McCoy et al. 2019)

arXiv: 1902.01007

2 of 7

What can you learn from title + abstract?

3 of 7

Kinds of NLP Papers

  • Theory!

“Prove transformers cannot learn to multiply arbitrary sequences in S5

  • Empiricism!

“Models rely on shallow heuristics to solve NLI tasks”

  • Task solving!

“Behold: the Transformer”

4 of 7

Empirical tests need benchmarks!

“Benchmark” = way to test on common ground

Good:

“Modifying model X by doing Y improves performance on Z”

Bad:

“Sentiment classification models work better on English than on Icelandic”

Why is the good example good?

Why is the bad example bad?

5 of 7

Sections & their purposes

From McCoy et al. (2019):

  1. Introduction
  2. Syntactic Heuristics
  3. Dataset Construction
  4. Experimental Setup
  5. Results
  6. Discussion
  7. Augmenting training data with HANS-like examples*
  8. Related Work
  9. Conclusion

Where do they state their hypothesis?

Why are (3) and (4) different sections?

What’s going on with (7)?

What’s missing?

What is their hypothesis?

6 of 7

Quantitative & qualitative explanations

7 of 7

Meta Questions

  • Why is this paper relevant?
  • Who is it written for?
  • Why are the authors writing it?
  • Why are you reading it?