https://github.com/buehlk/snowimputr

https://github.com/buehlk/snowimputr

Repository

README.Rmd

---
output: github_document
---



```{r, include = FALSE}
knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  fig.path = "man/figures/README-",
  out.width = "100%"
)
```

# Snowball bootstrap imputation of Telegram data with snowimputr

This package includes convenient functions to identify deleted messages in Telegram datasets using the consecutive numbering of message IDs when Telegram data is scraped via the Telethon API. Those IDs are then queried in the overall dataset to reconstruct the original message history of the dataset.

## Installation

You can install snowimputr from GitHub with:


```{r, message = FALSE}
# install.packages("devtools")
devtools::install_github("buehlk/snowimputr")

```

First, this package allows for quick checks about the extent of deleted messages in the Telegram dataset. The message IDs within a channel (or group) are always numbered consecutively, if scraped via Telethon. Therefore, missing numbers in a sequence mark a deleted message. This can also be very useful to make sense of possible gaps within the data. A Telegram dataset might look like this:


```{r, results = 'asis', warning=FALSE, message = FALSE}
library(snowimputr)
library(dplyr)

telegram_data <- data.frame(
 Channel_ID = rep(c("111111", "222222", "333333"), each = 6),
 Message_ID = c(c(1:4, 6, 7), c(2:7), c(1, 5, 9, 15, 16, 19)),
 Date = as.Date(c("2023-06-26", "2023-06-27", "2023-06-28", "2023-06-29", "2023-06-30", "2023-07-01",
                  "2023-06-27", "2023-06-28", "2023-06-29", "2023-06-29", "2023-06-30", "2023-07-01",
                  "2023-06-06", "2023-06-07", "2023-06-20", "2023-06-21", "2023-06-30", "2023-07-01"), format = "%Y-%m-%d"),
 Forwarded_from_channel_ID = c(c("333333", "333333", "222222", rep(NA, 3)),
                               c(rep(NA, 4), "111111", "333333"), rep(NA, 6)),
 Forwarded_from_message_ID = c(2, 3, 1, rep(NA, 7), 5, 2, rep(NA, 6)),
 Message_content = c(
   c(rep("Forwarded. Original message is deleted, but available via snowball bootstrap imputation.", 3), rep("This is an organic message.", 3)),
   c(rep("This is an organic message.", 4), rep("Forwarded. Original message is deleted, but available via snowball bootstrap imputation.", 2)),
   rep("This is an organic message.", 6)
 )
 )

flextable::flextable(telegram_data) %>% 
  flextable::align(align = "center") %>%
  flextable::width(j = 6, width = 7, unit = "cm") %>% 
  flextable::fontsize(15) %>% 
  plot()

```

To show all deleted messages in a Telegram dataset, use *list_deleted_msg*:
```{r results = 'asis', warning=FALSE}
 # Show all deleted messages in dataset:
deleted <- list_deleted_msg(msg_df = telegram_data,
 channel_ID_colname = "Channel_ID",
 message_ID_colname = "Message_ID")
 
flextable::flextable(deleted) %>%  
  flextable::align(align = "center") %>% 
  plot()


```

If one is only interested in the missing messages of one particular channel, this can also be stated in the *list_deleted_msg* function:
```{r results = 'asis', warning=FALSE}
 # Show the deleted messages of Channel "333333"
deleted2 <- list_deleted_msg(msg_df = telegram_data,
 channel_ID_colname = "Channel_ID",
 message_ID_colname = "Message_ID",
 target_entity = "333333")

flextable::flextable(deleted2) %>%  
  flextable::align(align = "center") %>% 
  plot()
```

Of course, deleted messages might contain important information. Buehling (2023) shows that missing of Telegram messages impair computational analyses of Telegram data. Also, deleted messages might be deleted for strategic reasons, which makes their contents even more interesting to researchers. If one is interested in the spread of information, then missing messages might obscure the source of an information! To find and impute those messages, the dataset is queried for forwarded messages in all channels and groups of the dataset. Metadata scraped via Telethon contains information on the channel a message is forwarded from, as well as the message ID it had in this channel. This way, one is able to reconstruct previously deleted messages. The querying can be done using *find_deleted_msg*:

```{r results = 'asis', warning=FALSE}
# Find the messages in the dataset that were deleted but forwarded in other channels
find_all <- find_deleted_msg(
 corpus_df = telegram_data,
 deleted_msg_ids = deleted$Message_ID,
 deleted_channel_ids = deleted$Channel_ID,
 fwd_from_colname = "Forwarded_from_channel_ID",
 fwd_msg_colname = "Forwarded_from_message_ID"
 )

flextable::flextable(find_all) %>%  
  flextable::align(align = "center") %>% 
  plot()
```

If you are again only interested in one particular channel, use:
```{r results = 'asis', warning=FALSE}
# Find the messages in the dataset that were deleted by "333333" but forwarded in other channels
find_particular <- find_deleted_msg(
 corpus_df = telegram_data,
 deleted_msg_ids = deleted2$Message_ID,
 deleted_channel_ids = deleted2$Channel_ID,
 fwd_from_colname = "Forwarded_from_channel_ID",
 fwd_msg_colname = "Forwarded_from_message_ID"
 )

flextable::flextable(find_particular) %>%  
  flextable::align(align = "center") %>% 
  plot()
```
You can see that the message ID 2 from channel "333333" has been forwarded to two different channels and is therefore found twice. You can easily choos to only show the "oldest" or "latest" forward:

```{r results = 'asis', warning=FALSE}
# Find the messages in the dataset that were deleted by "333333" but forwarded in other channels
find_oldest <- find_deleted_msg(
 corpus_df = telegram_data,
 deleted_msg_ids = deleted2$Message_ID,
 deleted_channel_ids = deleted2$Channel_ID,
 fwd_from_colname = "Forwarded_from_channel_ID",
 fwd_msg_colname = "Forwarded_from_message_ID",
 return_result = "oldest", date_colname = "Date"
 )

flextable::flextable(find_oldest) %>%  
  flextable::align(align = "center") %>% 
  plot()
```

Also, this package lets you conveniently substitute the message ID column and the channel ID column in the results table for the *original* channel ID and message ID, so that the resulting table can be conveniently appended to the original dataset for further analysis!


```{r  results = 'asis', warning=FALSE}
# Find the messages in the dataset that were deleted by "333333" but forwarded in other channels
find_oldest_impute <- find_deleted_msg(
 corpus_df = telegram_data,
 deleted_msg_ids = deleted2$Message_ID,
 deleted_channel_ids = deleted2$Channel_ID,
 fwd_from_colname = "Forwarded_from_channel_ID",
 fwd_msg_colname = "Forwarded_from_message_ID",
 return_result = "oldest", date_colname = "Date",
 message_ID_colname = "Message_ID",
 channel_ID_colname = "Channel_ID",
 impute = TRUE
 )

flextable::flextable(find_oldest_impute) %>%  
  flextable::align(align = "center") %>% 
  plot()
```

For more information about the effects of deleted Telegram messages on the results of computational analyses, see: 

Buehling, K. (2023). Message Deletion on Telegram: Affected Data Types and Implications for Computational Analysis. *Communication Methods and Measures*, 1-23.

Owner

GitHub Events