WhatsR

DOI R-CMD-check CRAN status Downloads TotalDownloads

WhatsR Sticker

This is an R-package to import exported WhatsApp chatlogs, parse them into a usable dataframe format and thereby enable further analysis. This parser was built with the goal to work with chat logs extracted on Android as well as iOS devices, run on Linux, Mac and Windows, and to be able to handle multiple languages. Currently, only English and German are supported, but in principle, other languages could be added relatively easily (see below). The repo also contains a function to scrape and update the emoji_dictionary, should new emoji be added to WhatsApp in the meantime.

How to set it up?

1) Requirements

  • If on Windows, RTools needs to be installed and working
  • The RJava package needs to be installed and working
  • plotting emoji requires the ragg package and you need to set your graphics backend to AGG (In Rstudio: Tools > Global Options > Graphics > Backend)

2) Installing it 

 # for the most up-to-date GitHub version
 library(devtools)
 devtools::install_github("gesiscss/WhatsR")
 
 # from CRAN
 install.packages("WhatsR")

3) Testing it 

 # creating simulated chatlog (saved in working directory)
 simulated_raw_chat <- create_chatlog(language = "english")
 
 # parsing it
 simulated_parsed_chat <- parse_chat("Simulated_WhatsR_chatlog.txt")
 
 # plotting emojis contained in chat
 plot_emoji(simulated_parsed_chat, plot="bar")

4) Using it with your own data

Extract chat from your phone 

 # parsing it
 simulated_parse_chat <- parse_chat("PATH/TO/YOUR/EXPORTED/FILE.txt")
 
  # plotting it
 plot_emoji(simulated_parse_chat, plot="bar")

Scientific use

If you are using this package for your research, please cite it accordingly. You get the citation as a BibTex by running

citation("WhatsR")
To cite package ‘WhatsR’ in publications use:

  Kohne J (2023). “WhatsR - An R-package for parsing, anonymizing and visualizing exported
  WhatsApp chat logs.” doi:10.5281/zenodo.7875622, <https://doi.org/10.5281/zenodo.7875622>.

A BibTeX entry for LaTeX users is

  @Misc{,
    title = {WhatsR - An R-package for parsing, anonymizing and visualizing exported WhatsApp chat logs},
    doi = {10.5281/zenodo.7875622},
    url = {https://doi.org/10.5281/zenodo.7875622},
    year = {2023},
    author = {J. Kohne},
  }

Does this parser work with other languages too?

Currently, only chats exported from phones set to German or English are supported. Other languages can be added by appending the languages.csv file with the necessary regular expressions to differentiate system messages from user generated content. In addition, parse_chat would need to be adapted and additional tests would have to be added. If you would like to add a language, please consider doing so via a pull request in this repository.

Examples

The package also includes some functions to compute additional metrics and visualize them. We will provide some basic examples for chats with two participants and for group chats with multiple participants here, for a complete overview, you can check the documentation or the figure section. The used chat is a chat that was parsed with the anonimize = TRUE parameter to exclude participant names. All plotting functions include multiple types of plots and additional parameters to restrict the range of the data.

Token Summary per Person 

summarize_tokens_per_person(data)
$`WhatsApp System Message`
$`WhatsApp System Message`$Timespan
$`WhatsApp System Message`$Timespan$Start
[1] "2020-10-27 18:51:00 UTC"

$`WhatsApp System Message`$Timespan$End
[1] "2022-10-06 19:57:00 UTC"


$`WhatsApp System Message`$TokenStats
   Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
      1       1       1       1       1       1 


$Person_1
$Person_1$Timespan
$Person_1$Timespan$Start
[1] "2020-10-27 18:51:00 UTC"

$Person_1$Timespan$End
[1] "2022-10-06 19:57:00 UTC"


$Person_1$TokenStats
   Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
  1.000   1.000   6.000   9.195  13.000 169.000 


$Person_2
$Person_2$Timespan
$Person_2$Timespan$Start
[1] "2020-10-27 18:51:00 UTC"

$Person_2$Timespan$End
[1] "2022-10-06 19:57:00 UTC"


$Person_2$TokenStats
   Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
   1.00    1.00    6.00   10.75   14.00  407.00

Message Distribution

Distribution of sent Messages.

plot_messages(data, plot = "cumsum", exclude_sm = TRUE)

Token Distribution

Distribution of sent Tokens (words).

plot_tokens(data, plot = "violin", exclude_sm = TRUE)

Tokens over Time

Distribution of sent Tokens per Person over time

plot_tokens_over_time(data, plot = "hours", exclude_sm = TRUE)

Wordcloud

Wordcloud of sent tokens, seperately for each participant.

plot_wordcloud(data, comparison = TRUE, exclude_sm = TRUE, font_size=50, min_occur= 300)

Lexical Dispersion Plot

Occurrences of keywords in the chat. Example keyword is “Weihnachten” (Christmas).

plot_lexical_dispersion(data,keywords = c("weihnachten"), exclude_sm = TRUE)

Amount of sent Links per person and over time

plot_links(data, plot = "heatmap", exclude_sm = TRUE)

Sent Smilies

Amount of sent Smilies over time

plot_smilies(data, plot = "cumsum", exclude_sm = TRUE)

Sent Emoji

Amount of sent emoji per person

plot_emoji(data, plot = "splitbar", min_occur = 50, exclude_sm = TRUE)

Location Visualization [Temporarily disabled]

Plotting mentioned locations by persons

Currently disabled until changes in ggmap make it to CRAN

Replytimes

Plotting time it takes to respond

plot_replytimes(data, type = "replytime", exclude_sm = TRUE)

Sent Media

Amount of sent Media files per person and over time

plot_media(data, plot = "bar", exclude_sm = TRUE)

Interactive Networks

Interactive network of chat participants. A connection represents a response to a message. Each Message is interpreted as a response to the previous message. Consecutive messages by the same chat participant are summarized into one “session”. The shown plot is simple image, the actual output is an interactive HTML object, see man folder.

plot_network(data)