/babble_corpus

Primary LanguageR

babble_corpus

data/

contains data for these scripts:

  • list of clips per part
  • answers spreadsheets
  • linking file (metadata to clips)

babble_analysis_basic

Retrieves

  • number of clips with at least one annotation
  • number of clips with at least one annotation missing (only works for part 2 and 3 where 3 annotations were required)
  • number of clips with at least 3 annotations
  • number of clips with 100% agreement
  • number of cips without majority agreement (only works for part 2 and 3 where majority=>=2)

check_majority_agreement

check majority agreement on a preprocessed csv (see Rmd itself) and writes csv with clips without majority agreement

babble_part_metadata

from linking file and answer spreadsheet, retrieves:

  • number of children by corpus in this part
  • number of utterances by child
  • number of clips by child
  • total number of clips in this part
  • histogram of the length of utterances in this part

Majority agreement

Script to determine the minimum sample size needed to obtain a stable majority agreement. It requires:

  • A data file with at least 20k entries (the chunks indeces are hard-coded, but can be changed if the file is smaller) It returns, for every chunk:
  • number of clips with at least one annotation
  • number of clips with 2 or more annotations
  • the proportion of majority agreement for the given data (includes both 2/3 agreement and 3/3 agreement)