Tools to Perform ‘Misinformation’ Analysis on a Text Corpus
Tools to Perform ‘Misinformation’ Analysis on a Text Corpus
The following functions are implemented:
mi_analyze_document: Run a one or more documents through a misinformation/bias sentiment analysismi_plot_document_summary: Plot raw frequency count summary from a processed corpusmi_plot_individual_frequency: Plot raw frequency count summary from a processed corpusmi_use_builtin: Use a built-in ‘sentiment’ word/phrase listword_doc_to_txt: Convert a Word document to a plaintext document
devtools::install_github("hrbrmstr/misinfo")library(misinfo)
library(hrbrthemes)
library(tidyverse)
# current verison
packageVersion("misinfo")## [1] '0.1.0'
mi_analyze_document(
path = system.file("extdat", package="misinfo"),
pattern = ".*txt$",
bias_type = "explanatory",
sentiment_list = mi_use_builtin("explanatory")
) -> corpusglimpse(corpus)## List of 4
## $ bias_type : chr "explanatory"
## $ corpus :Classes 'tbl_df', 'tbl' and 'data.frame': 7 obs. of 3 variables:
## ..$ doc : chr [1:7] "CNET - How US cybersleuths decided Russia hacked the DNC" "CNN - DNC hack" "IVN - CNN Covering For DNC Corruption" "Red Nation- FBI Didn't Analyze DNC servers" ...
## ..$ text : chr [1:7] "how us cybersleuths decided russia hacked the dnc digital clues led security pros to agencies in putins govern"| __truncated__ "dnc hack what you need to know by tal kopan cnn updated 130 pm et tue june 21 2016 cyber experts russia hack"| __truncated__ "cnn covering for dnc corruption by jeff powers in opinion jun 27 2017 we should have known this hysteria woul"| __truncated__ "fbi didnt analyze dnc servers before issuing hacking report posted at 839 pm on january 4 2017 by jennifer van"| __truncated__ ...
## ..$ doc_id: chr [1:7] "CNET - How US cybersleuths dec" "CNN - DNC hack" "IVN - CNN Covering For DNC Cor" "Red Nation- FBI Didn't Analyze" ...
## $ frequency_summary :Classes 'tbl_df', 'tbl' and 'data.frame': 7 obs. of 5 variables:
## ..$ doc_id : chr [1:7] "CNET - How US cybersleuths dec" "CNN - DNC hack" "IVN - CNN Covering For DNC Cor" "Red Nation- FBI Didn't Analyze" ...
## ..$ n : int [1:7] 12 3 3 2 33 6 13
## ..$ doc_num : chr [1:7] "1" "2" "3" "4" ...
## ..$ total_words: int [1:7] 1327 1024 880 377 5119 1430 1694
## ..$ pct : num [1:7] 0.00904 0.00293 0.00341 0.00531 0.00645 ...
## $ individual_frequency:Classes 'tbl_df', 'tbl' and 'data.frame': 119 obs. of 6 variables:
## ..$ doc_id : chr [1:119] "CNET - How US cybersleuths dec" "CNN - DNC hack" "IVN - CNN Covering For DNC Cor" "Red Nation- FBI Didn't Analyze" ...
## ..$ keyword : chr [1:119] "because" "because" "because" "because" ...
## ..$ ct : int [1:119] 1 0 2 0 2 2 2 0 0 0 ...
## ..$ doc_num : chr [1:119] "1" "2" "3" "4" ...
## ..$ total_words: int [1:119] 1327 1024 880 377 5119 1430 1694 1327 1024 880 ...
## ..$ pct : num [1:119] 0.000754 0 0.002273 0 0.000391 ...
## - attr(*, "class")= chr "misinfo_corpus"
corpus$frequency_summary## # A tibble: 7 x 5
## doc_id n doc_num total_words pct
## <chr> <int> <chr> <int> <dbl>
## 1 CNET - How US cybersleuths dec 12 1 1327 0.00904
## 2 CNN - DNC hack 3 2 1024 0.00293
## 3 IVN - CNN Covering For DNC Cor 3 3 880 0.00341
## 4 Red Nation- FBI Didn't Analyze 2 4 377 0.00531
## 5 The Nation - A New Report Rais 33 5 5119 0.00645
## 6 VOA- Think Tank- Cyber Firm at 6 6 1430 0.00420
## 7 WaPo - DNC hack 13 7 1694 0.00767
corpus$individual_frequency## # A tibble: 119 x 6
## doc_id keyword ct doc_num total_words pct
## <chr> <chr> <int> <chr> <int> <dbl>
## 1 CNET - How US cybersleuths dec because 1 1 1327 0.000754
## 2 CNN - DNC hack because 0 2 1024 0
## 3 IVN - CNN Covering For DNC Cor because 2 3 880 0.00227
## 4 Red Nation- FBI Didn't Analyze because 0 4 377 0
## 5 The Nation - A New Report Rais because 2 5 5119 0.000391
## 6 VOA- Think Tank- Cyber Firm at because 2 6 1430 0.00140
## 7 WaPo - DNC hack because 2 7 1694 0.00118
## 8 CNET - How US cybersleuths dec therefore 0 1 1327 0
## 9 CNN - DNC hack therefore 0 2 1024 0
## 10 IVN - CNN Covering For DNC Cor therefore 0 3 880 0
## # ... with 109 more rows
mi_plot_individual_frequency(corpus)
mi_plot_individual_frequency(corpus, FALSE)
mi_plot_document_summary(corpus)
mi_plot_document_summary(corpus, FALSE)
map_df(c("uncertainty", "sourcing", "retractors", "explanatory"), ~{
# compute individual bias scores
mi_analyze_document(
path = system.file("extdat", package="misinfo"),
pattern = ".*txt$",
bias_type = .x,
sentiment_list = mi_use_builtin(.x)
) -> corpus
mutate(corpus$frequency_summary, bias = corpus$bias_type)
}) -> biases
mutate(biases, pct = ifelse(pct == 0, NA, pct)) %>%
ggplot(aes(sprintf("Doc #%s", doc_num), bias, fill=pct)) +
geom_tile(color="#2b2b2b", size=0.125) +
scale_x_discrete(expand=c(0,0)) +
scale_y_discrete(expand=c(0,0)) +
viridis::scale_fill_viridis(direction=-1, na.value="white") +
labs(x=NULL, y=NULL, title="Word List Usage Heatmap (normalized)") +
theme_ipsum_rc(grid="")
distinct(biases, doc_num, doc_id, total_words) %>%
knitr::kable(format="markdown")| doc_id | doc_num | total_words |
|---|---|---|
| CNET - How US cybersleuths dec | 1 | 1327 |
| CNN - DNC hack | 2 | 1024 |
| IVN - CNN Covering For DNC Cor | 3 | 880 |
| Red Nation- FBI Didn’t Analyze | 4 | 377 |
| The Nation - A New Report Rais | 5 | 5119 |
| VOA- Think Tank- Cyber Firm at | 6 | 1430 |
| WaPo - DNC hack | 7 | 1694 |