emnlp camera-ready revisions

[hawkrobe]

• [x] add new section to describe switchboard methods/analyses
  • discuss pros and cons of natural data vs. judgements at beginning paragraph of this section
• [x] go through response letter and act on the things we mentioned
  • [x] clarify in intro aim of paper
  • [x] clarify/soften stance on gradience (move discussion of continuous DV out of supplemental methods)
  • [x] give examples? (in supplemental?)
  • [x] cite more psycholinguistics in relevant work/discussion (e.g. Filik & Paterson)
• [x] move layer-wise analysis of gpt-2 into main text (supplemental figure)
• [x] cite those recent cogsci papers in relevant work (check old email to adele and scroll up in slack channel)