To encourage further investigation of trigger-action programming in the real world, we have released our CHI '16 dataset of 224,590 IFTTT recipes (as of September 2015) to fellow academic researchers. Per our agreement with IFTTT, this dataset may only be used for academic, non-commercial purposes. Note that you may not re-distribute the raw dataset. You may, however, distribute aggregate findings, such as writing an academic paper or blog post with the results of your analyses.
To access the dataset, you first need to email Blase (blase at blaseur.com) with a statement that you will only use this dataset for academic, non-commercial purposes. He will reply to you with a username and password, which you can then use to access the data files.
We are distributing the data in a tab-separated file (tsv). The 11 fields are separated by tabs, while recipes are separated by newlines. To get you started analyzing the dataset, we have written a sample script for R that reads in the data and identifies the three most prolific authors. Once you have downloaded our dataset, also download our sample script to the same directory. You can run it from the command line as follows: R CMD BATCH iftttexample.R
One example recipe from our data, with the column names in bold, follows: