Repository navigation
Confusing colors on the textplot_keyness plot. #1233
Description
Activity
Thanks for the feedback. @koheiw and I are in the process of fixing this plot, and will add to #1211 the issue about the words. I think they should be black by default. Also, when there just two classes, we will have the legend label read the name of the reference document, rather than the generic "Reference".
Yes, you should be able to change the plot using ggplot2
+additions.Awesome! Off topic, thank you for building this package. It's truly powerful and has been the basis of building out my companies text analytics platform! I will be posting all the code in my GitHub!
Two issues we have now discovered, in the course of investigating this:
- Colors for text labels can be flipped, as in the OP example, or as in:
corpus_subset(data_corpus_inaugural, President %in% c("Obama", "Trump")) %>% dfm(remove = stopwords("english"), remove_punct = TRUE) %>% textstat_keyness(target = "2017-Trump") %>% textplot_keyness()
-
For a dfm with just two documents, we should label the reference document as the doc name in the legend, not the more generic "Reference". (The previous code illustrates this problem.)
-
In the legend, the target group should be first. It is currently alphabetical.
corpus_subset(data_corpus_inaugural, President %in% c("Obama", "Trump")) %>% dfm(remove = stopwords("english"), remove_punct = TRUE, groups = "President") %>% textstat_keyness(target = "Trump") %>% textplot_keyness()
when plotting keyness, the colors used are red and blue, blue being the target. The target bars are blue but the text is red. Why the difference? Also, can we pass ggplot arguments to textplot to alter the style?