Plotting Bigrams in Bar Chart with ggplot2

Question

My data looks like this:

> str(bigrams_joined)
Classes ‘tbl_df’, ‘tbl’ and 'data.frame':   71319 obs. of  2 variables:
 $ line   : int  1 1 1 1 1 1 1 1 1 1 ...
 $ bigrams: chr  "in practice" "practice risk" "risk management" "management is"

I would like to plot the top 10 or 15 most frequently occurring bigrams in my dataset to a bar chart in ggplot2 and have the bars running horizontally with the labels on the y-axis.

Any help with this is greatly appreciated!

Thank you

Top 15 *most frequently* occurring - sorry I didn't specify. Editing now. — Davide Lorino, Apr 20 '18 at 11:18
this? `ggplot(bigrams_joined, aes(bigrams)) + stat_count(geom="bar")` — Roman, Apr 20 '18 at 11:26
Thanks for your suggestion - unfortunately that tries to plot every single value, rather than the top 10 or 15. I have about 35000 rows and i'm only interested in the top 10 or 15 — Davide Lorino, Apr 20 '18 at 11:30

phiver · Answer 1 · 2018-04-20T11:47:24.947

1

You could something like this, dplyr's top_n function to filter the top 15 bigrams + ggplot to plot them.

library(dplyr)
library(ggplot2)


bigrams_joined %>%
  top_n(15, bigrams) %>% 
  ggplot(aes(bigrams)) + 
  geom_bar() +  
  coord_flip()

or ordered:

bigrams_joined %>%
  group_by(bigrams) %>% 
  mutate(n = n()) %>% 
  ungroup() %>% 
  top_n(15, bigrams) %>% 
  mutate(bigrams = reorder(bigrams, n)) %>%
  ggplot(aes(bigrams)) + 
  geom_bar() +
  coord_flip()

edited Apr 20 '18 at 11:47

answered Apr 20 '18 at 11:33

phiver

23,048
14
44
56

Thank you for your suggestion, unfortunately it is returning the top 15 sorted in alphabetical order from Z to A. I'm not sure why that is the convention of top_n() but that's what it is doing in this case. – Davide Lorino Apr 20 '18 at 12:05
that is not the top_n function, but ggplot taking over. – phiver Apr 20 '18 at 12:28
Ahh I see - thank you for your help with this though – Davide Lorino Apr 20 '18 at 12:33

score 1 · Accepted Answer · answered Apr 22 '18 at 20:58

Looks like you need to count() your bigrams (from dplyr), and then you need to order them in your plot. For that these days, I prefer to use something like fct_reorder() from forcats.

library(janeaustenr)
library(tidyverse)
library(tidytext)

data_frame(txt = prideprejudice) %>%
    unnest_tokens(bigram, txt, token = "ngrams", n = 2) %>%
    count(bigram, sort = TRUE) %>%
    top_n(15) %>%
    ggplot(aes(fct_reorder(bigram, n), n)) +
    geom_col() +
    coord_flip() +
    labs(x = NULL)
#> Selecting by n

Created on 2018-04-22 by the reprex package (v0.2.0).

Thank you again Julia, another perfect answer! – Davide Lorino Apr 24 '18 at 03:05 — Davide Lorino, Apr 24 '18 at 03:05

Plotting Bigrams in Bar Chart with ggplot2

2 Answers2