4

I am looking for a dataset containing academic recommendation letters, written in English, with as many following fields as possible:

  • content of the letter
  • date of the letter
  • purpose of the letter (e.g., faculty application, visa application, internship application, etc.)
  • relationship between recommendee and recommender
  • demographics on recommendee and recommender
Franck Dernoncourt
  • 7,780
  • 9
  • 39
  • 86
  • 2
    How do you properly anonymise such letters? I doubt very much that this can be turned into Open Data. –  Feb 19 '16 at 10:02
  • @jknappen That's one of the elements I am curious about :) maybe http://softwarerecs.stackexchange.com/q/27264/903 (but for FERPA if in the US)+ manual checking. Also maybe more lenient outside the US. – Franck Dernoncourt Feb 19 '16 at 14:46

1 Answers1

2

Idea 1: Troll google for personal letters of reference with a search engine - here are some samples:

  • include the strings "letter of reference", "to whom it may concern"
  • exclude the string "template"
  • filter for PDF and DOC files

example query

examples results pdf1, pdf2, pdf3

enhanced query:

  • include degree (Master, PhD)

  • include department (Physics, Law)

The downside is you'll have to do the categorization yourself

Idea 2: (ethical?) Post fake job advertisements on Craiglist or Indeed and ask people to send CV with letters of reference. At least one would then have geographical and topic-based categorization.

Idea 3+: There are plenty of less-ethical things beyond here. That being said, analysis of the hiring process is an important research topic and leads to important discoveries like - Are Emily and Greg More Employable than Lakisha and Jamal? A Field Experiment on Labor Market Discrimination.

philshem
  • 17,647
  • 7
  • 68
  • 170
  • 1
    1 and 2 are not really suitable for academic letters. I'm also looking for letters, and would be happy with non-academic ones, too. 1 seems sort of reasonable, but a lot of work. Also, IRB approval is uncertain.... – Andreas Mueller Apr 13 '16 at 20:25
  • 2
    For mining google, it might be interesting to look at city council applications. They seem to be public in some cities, which could make the use legal. – Andreas Mueller Apr 13 '16 at 20:44