NOTE
This package is not being maintained.
ChildRecordsR is an R package that provides and facilitates the analysis of annotations (often from daylong recordings) formatted using ChildRecordsData. The main functions in this package are data aggregation and analysis of the reliability of (human or machine) annotators.
Before you start, make sure you have some data formatted using ChildRecordsData. A description of the format is provided here. Documentation concerning the setup and installation of corpora in ChildRecordsData format can be found here. Available corpora are listed here. See Appendix for an example.
In R terminal:
install.packages("devtools")
library(devtools)
install_github("LAAC-LSCP/ChildRecordsR")
- Checking the integrity of your ChildRecordsData folder.
- Finding times and files common to multiple annotators
- Aggregating of their annotations and their transformation to multiple formats
- Calculating indicators of annotation reliability
- Evaluating the annotations' quality with reliability and classification indicators
An R tutorial/vignette can be found here, which provides you a guide to the basic functions and analyses from the package. Member of the LAAC or who have restricted access to the Namibia data can use the Namibia tutorial here
At any point, you can also use the help in R and Rstudio, using ?function_name where function_name is the name of a function or in the help panel if you are using Rstudio. You can also found functions' documentation on the https://laac-lscp.github.io/ChildRecordsR.
If any issues arise, feel free to post them here
You can find the list of formatted LAAC datasets as well as instructions to get them here
All you need to do these analyses are the contents of annotations/. So typically, you'll need to:
- Install datalad
- install the dataset you want. Note that this just installs the structure of the dataset, but not the contents.
- Get the contents of annotations/ by doing, from within your local copy of the dataset,
datalad get annotations/.