This is the code for processing the 37K dataset presented in How do People Type on Mobile Devices? Observations from a Study with 37,000 Volunteers, currently support computing Word Modified Ratio (WMR) and Auto-Correction Ratio (AC).
For using the code, you need to download the dataset and put the csv files in the original_data folder.
For some reason the IKI recorded in the original open_input_test_sections can not be trusted 100%, we also provide function computing iki based on the TIMESTAMP
- --data-cleaning Generate the cleaned dataset
- --analyze Analyze the dataset and save the visaulization sheets
- --visualize Generate the plots and save them
- --visualize-by-edit-distance split the data by edit distance for visualization
- --visualize-by-sentence-length split the data by reference sentence length for visualization
- --keyboard The keyboard used by the participants, currently support "Gboard", without his option, the code will process all the keyboards
- --auto-correct Whether to compute the Auto-Correction Ratio (AC)
- --os Which OS is used when typing, supporting Android and iOS
- --modification Compute the modification ratio on characters level
- --wmr Compute the Word Modified Ratio (WMR) on word level
- --ac Compute the Auto-Correction Ratio (AC) on word level
- --edit-distance Compute the edit distance on character level
- --age get the age vs iki
- --num get the number of test section vs iki
- --filter Filter the dataset by the modification ratio, need data cleaning first (modify inside code for above or below)
python main.py --data-cleaning --auto-correct --keyboard Gboardpython main.py --filter 0.8 --auto-correct --keyboard Gboardpython main.py --filter --iki 345.5 --rebuildpython main.py --analyze --keyboard Gboard --auto-correct --modification --wmr --acpython main.py --visualize --keyboard Gboard --auto-correct --modification --wmr --ac