Open University Pair-Programming Dialogues

This is the dataset collected for the PhD project Automated processing of referring acts in pair-programming dialogue by Cecilia Domingo under the supervision of Paul Piwek, Michel Wermelinger, and Svetlana Stoyanchev. The PhD thesis describes how the data was collected and annotated, though the relevant documentation is also included here.The dataset contains 23 dialogues, with audio recordings, screen-capture videos, code files, keylog records, dialogue json files combining the transcripts with the code and keylog files, and reference annotations in Excel and json format. The dialogues were recorded between May-June 2023 (pilot phase) and March-April 2024 (main phase). The participants were Open University students who used Python in their course work; profiles are thus very varied. More details can be found on the thesis document.Please download all file parts into the same directory to be able to decompress them - when you use a zip processing software to decompress one part, it will automatically detect all the other parts if you have not changed their name.The dataset has been anonymised to protect participants' privacy. Demographic data has been discretised (e.g., age groups instead of specific ages). Names mentioned in writing have been deleted, and names spoken have been bleeped or silenced. Screenshots with any personal information (browser tabs, email address, pet pictures, etc.) have been deleted, cropped, or heavily blurred, and small items covered. Voice recordings have been anonymised using voice conversion for highest intelligibility; however, due to accent variations affecting the performance of voice conversion models, alternative (less natural but without glitches) anonymised recordings are included which have been anonymised using voice filters (including pitch shifting, modifications to selected equalisation bands, chorus modulation, and the addition of noise).