source("./device-setup.R")
library(tidyverse)The data may be downloaded from the project’s page: http://ns.inria.fr/loki/WordSuggestions/.
This table contains one row for each participant. The different columns are as described below:
This table contains three rows per participant providing data about each three parts of the experiment. Each part, called a “run” is dedicated to the use of one device in particular. The different columns are as described below:
Participant Identifier.
The word suggestion accuracy condition.
The device used for the run.
The order of the device in the experiment.
The start date and time of the run in ISO 8601 format.
The end date and time of the run in ISO 8601 format (only provided for the desktop condition).
The duration of the run in seconds.
The start date and time of the experiment in ISO 8601 format.
The end date and time of the experiment in ISO 8601 format.
The duration of the experiment in seconds.
The self-reported age of the participant.
The self-reported gender of the participant.
Answer for “In the last 7 days, in total, how long have you swipe typed?” This was provided once per experiment.
Answer for “In the last 24 hours, how many word suggestions have you used when typing on [device]?” This was asked for laptop, tablet and phone at the beginning of the experiment.
Answer for “In the last 7 days, in total, how long have you used [device]?” This was asked for laptop, tablet and phone at the beginning of the experiment.
Answer for “In the last 7 days, in total, how long have you typed on [device] (using any typing method)?” This was asked for laptop, tablet and phone at the beginning of the experiment.
Answer for “In the last 7 days, in total, how long have you typed on a phone with only the thumb of one hand?” This is only available for the phone condition.
Agreement for “The controls (keyboard and word suggestions) are satisfactory for the completion of the task.” This was provided for each condition.
Agreement for “The word suggestions are accurate.” This was provided for each condition.
Agreement for “The use of the keyboard is efficient in this task.” This was provided for each condition.
Agreement for “The word suggestions are distracting.” This was provided for each condition.
NASA-TLX mental demand scale. This was provided for each condition.
NASA-TLX physical demand scale. This was provided for each condition.
NASA-TLX temporal demand scale. This was provided for each condition.
NASA-TLX performance scale. This was provided for each condition.
NASA-TLX performance scale. This was provided for each condition.
NASA-TLX effort scale. This was provided for each condition.
The average typing speed of the participant (without suggestions) in character per seconds.
The average typing speed of the participant (without suggestions) in
words per minutes (avg_cps * 60 / 5).
The minimum delay before updating suggestions. This is always 150ms.
The number of suggestions showed to the participant. This is always 3.
Register the number of trials during which the page was unfocused, which is an indication that the participant may be distracted.
The number of trials during which at least one suggestion took more than 300ms before being udpated.
TRUE if the experiment was completed.
Some runs were not included in our analysis. The column
run_rejection_reason provides the reason for the
exclusion.
The reason for excluding the run from the analysis. This is only provided for runs that were excluded.
read_device_runs(measured_only = FALSE)This table contains data logs from all participant trials.
Participant Identifier.
The device used for the run.
The word suggestion accuracy condition.
The number of the trial.
The identifier of the trial.
The phrase participant had to type during the trial.
If the trial was part of a practice block.
The number of character in the phrase to type.
The maximum number of keystrokes that could be saved using word suggestions during the trial.
The start date and time of the trial in ISO 8601 format.
The end date and time of the trial in ISO 8601 format.
The duration of the trial in seconds.
The number of key that were pressed to complete the trial.
The number of keystrokes that were saved using word suggestions. This may be negative if incorrect suggestions were used.
The number of suggestions that were used during the trial.
The number of incorrect suggestions that were used during the trial.
The number of characters that were manually entered then deleted by participant during the trial.
The number of characters that were inserted from a suggestion then deleted by participant during the trial.
The number of characters from the final participant input that were manually inserted.
The number of characters that were inserted during the trial either from suggestion or manual input (some may have been removed later).
TRUE if incorrect characters were inserted at some point during the trial (they had to be removed later to validate the trial).
The date and time where the first character was inserted during the trial (ISO 8601 format).
The date and time where the phrase was fully completed and correct for the first time (ISO 8601 format). Note that this is before the trial was validated.
The time it took for the participant to complete the phrase in seconds.
The average time it took for suggestions to be updated during the trial.
The standard deviation of the time it took for suggestions to be updated during the trial.
The number of times the experiment web page was unfocused during the trial.
TRUE if total_blur_events > 0
The maximum key stroke saving of the trial. This should be close from
accuracy.
The key stroke saving excluding errors.
The key stroke saving.
Entry speed in characters per second
(i.e. (total_chars - 1) / trial_duration).
If at least one suggestion set took more than 300ms to update.
If the run including this trial was included in our statistical analysis.
If this trial was included in our statistical analysis. See below for reasons to exclude a trials from our analysis.
The reason for excluding the trial from the analysis. This is only provided for trials that were excluded.
trials <- read_device_trials(measured_only = FALSE)
trialsThis large table contains all event logs recorded during our experiment. It includes the following columns:
The participant identifier.
The accuracy condition.
The device condition.
The trial identifier.
The number of the event in the trial.
The type of the event. INIT: trial initialization, UPDATE_SUGGESTIONS: update of the suggestions, INPUT_CHAR: manual insertion of a character by the participant at the end of their input, DELETE_CHAR: removal of the last input character by the participant, INPUT_SUGGESTION: insertion of a suggestion.
The content of participant’s input after the event.
If there is no errors in the input after the event.
The characters added to the input as a result of the event.
The characters removed from the input as a result of the event.
The number of characters added as result of the event (may be negative).
The number of character left to complete the phrase.
If the phrase is completed.
For events resulting from a web request (in particular updates of word suggestions), the moment the request was sent to the server (ISO 8601 format).
For events resulting from a web request (in particular updates of word suggestions), the moment the response from the server was received (ISO 8601 format).
The date and time of the event (ISO 8601 format).
Duration between this event and the previous one.
If the run including this event was included in our statistical analysis.
If the trial including this event was included in our statistical analysis.
If this event was included in our statistical analysis.
The reason for excluding the event from the analysis. This is only provided for events that were excluded.
The next word to type or the word currently being typed.
The number of the target word in the phrase.
trials %>% filter(is_run_measured & is_practice == F) %>% count(is_trial_measured)trials %>% filter(is_trial_measured) %>%
count(does_trial_have_errors)The following trials were removed because the participant was interrupted (the page was unfocused).
trials %>% filter(is_run_measured & is_practice == F) %>%
count(was_trial_blurred)trials %>% filter(is_run_measured & is_practice == F) %>%
count(has_long_suggestion_delay)