Help with an error when using a Tidy Tuesday example

There is a fantastic Tidy Tuesday blog post by Julia Silge in October of 2022 called " Find high FREX and high lift words for #TidyTuesday Stranger Things dialogue". The code is all there, and in a video Julia walks through creating and using the script. All excellent. My problem is that when I copy and run the code in Rstudio I get an error about using "frex" with matrix =. I can't seem to track down a solution. Here is all the code, up to and including where the error occurs.

Does anyone have an idea of why this works in the video but not when I try to run in Rstudio? Appreciate any help becuase this looks like a great addition to text mining. Thanks Julia, and thanks for any suggestions.

library(tidyverse)
episodes_raw <- read_csv('https://raw.githubusercontent.com/rfordatascience/tidytuesday/master/data/2022/2022-10-18/stranger_things_all_dialogue.csv')
#> Rows: 32519 Columns: 8
#> ── Column specification ────────────────────────────────────────────────────────
#> Delimiter: ","
#> chr  (3): raw_text, stage_direction, dialogue
#> dbl  (3): season, episode, line
#> time (2): start_time, end_time
#> 
#> ℹ Use `spec()` to retrieve the full column specification for this data.
#> ℹ Specify the column types or set `show_col_types = FALSE` to quiet this message.

dialogue <-
  episodes_raw %>%
  filter(!is.na(dialogue)) %>%
  mutate(season = paste0("season", season))

dialogue
#> # A tibble: 26,041 × 8
#>    season  episode  line raw_text                stage…¹ dialo…² start…³ end_t…⁴
#>    <chr>     <dbl> <dbl> <chr>                   <chr>   <chr>   <time>  <time> 
#>  1 season1       1     9 [Mike] Something is co… [Mike]  Someth… 01'44"  01'48" 
#>  2 season1       1    10 A shadow grows on the … <NA>    A shad… 01'48"  01'52" 
#>  3 season1       1    11 -It is almost here. -W… <NA>    It is … 01'52"  01'54" 
#>  4 season1       1    12 What if it's the Demog… <NA>    What i… 01'54"  01'56" 
#>  5 season1       1    13 Oh, Jesus, we're so sc… <NA>    Oh, Je… 01'56"  01'59" 
#>  6 season1       1    14 It's not the Demogorgo… <NA>    It's n… 01'59"  02'00" 
#>  7 season1       1    15 An army of troglodytes… <NA>    An arm… 02'00"  02'02" 
#>  8 season1       1    16 -Troglodytes? -Told ya… [chuck… Troglo… 02'02"  02'05" 
#>  9 season1       1    18 [softly] Wait a minute. [softl… Wait a… 02'08"  02'09" 
#> 10 season1       1    19 Did you hear that?      <NA>    Did yo… 02'10"  02'12" 
#> # … with 26,031 more rows, and abbreviated variable names ¹​stage_direction,
#> #   ²​dialogue, ³​start_time, ⁴​end_time

#____

library(tidytext)

tidy_dialogue <-
  dialogue %>%
  unnest_tokens(word, dialogue)

tidy_dialogue
#> # A tibble: 143,885 × 8
#>    season  episode  line raw_text                  stage…¹ start…² end_t…³ word 
#>    <chr>     <dbl> <dbl> <chr>                     <chr>   <time>  <time>  <chr>
#>  1 season1       1     9 [Mike] Something is comi… [Mike]  01'44"  01'48"  some…
#>  2 season1       1     9 [Mike] Something is comi… [Mike]  01'44"  01'48"  is   
#>  3 season1       1     9 [Mike] Something is comi… [Mike]  01'44"  01'48"  comi…
#>  4 season1       1     9 [Mike] Something is comi… [Mike]  01'44"  01'48"  some…
#>  5 season1       1     9 [Mike] Something is comi… [Mike]  01'44"  01'48"  hung…
#>  6 season1       1     9 [Mike] Something is comi… [Mike]  01'44"  01'48"  for  
#>  7 season1       1     9 [Mike] Something is comi… [Mike]  01'44"  01'48"  blood
#>  8 season1       1    10 A shadow grows on the wa… <NA>    01'48"  01'52"  a    
#>  9 season1       1    10 A shadow grows on the wa… <NA>    01'48"  01'52"  shad…
#> 10 season1       1    10 A shadow grows on the wa… <NA>    01'48"  01'52"  grows
#> # … with 143,875 more rows, and abbreviated variable names ¹​stage_direction,
#> #   ²​start_time, ³​end_time

#______

library(tidylo)

tidy_dialogue %>%
  count(season, word, sort = TRUE) %>%
  bind_log_odds(season, word, n) %>%
  filter(n > 20) %>%
  group_by(season) %>%
  slice_max(log_odds_weighted, n = 10) %>%
  mutate(word = reorder_within(word, log_odds_weighted, season)) %>%
  ggplot(aes(log_odds_weighted, word, fill = season)) +
  geom_col(show.legend = FALSE) +
  facet_wrap(vars(season), scales = "free") +
  scale_y_reordered() +
  labs(y = NULL)


#______

dialogue_sparse <-
  tidy_dialogue %>%
  mutate(document = paste(season, episode, sep = "_")) %>%
  count(document, word) %>%
  filter(n > 5) %>%
  cast_sparse(document, word, n)

dim(dialogue_sparse)
#> [1]  34 562

#_________

library(stm)
#> stm v1.3.6 successfully loaded. See ?stm for help. 
#>  Papers, resources, and other materials at structuraltopicmodel.com
set.seed(123)
topic_model <- stm(dialogue_sparse, K = 5, verbose = FALSE)

summary(topic_model)
#> A topic model with 5 topics, 34 documents and a 562 word dictionary.
#> Topic 1 Top Words:
#>       Highest Prob: you, i, the, to, a, and, it 
#>       FREX: max, mean, they're, i'm, don't, i, know 
#>       Lift: dart, clarke, mr, soon, better, duck, mistakes 
#>       Score: girlfriend, max, dart, duck, mr, building, kline's 
#> Topic 2 Top Words:
#>       Highest Prob: you, i, the, a, to, it, and 
#>       FREX: he's, let, we, he, go, us, what 
#>       Lift: flayer, party, fact, flayed, children, hold, tied 
#>       Score: flayer, ice, cherry, says, bob, key, code 
#> Topic 3 Top Words:
#>       Highest Prob: you, i, the, to, a, and, that 
#>       FREX: eddie, as, only, chrissy, make, has, much 
#>       Lift: ray, being, boobies, california, dad, deal, death 
#>       Score: ray, eddie, only, chrissy, try, had, vecna 
#> Topic 4 Top Words:
#>       Highest Prob: you, i, the, to, it, a, and 
#>       FREX: go, mike, come, jonathan, on, okay, get 
#>       Lift: jonathan, gone, jingle, kids, terry, answer, blood 
#>       Score: christmas, jonathan, copy, jingle, gone, bell, scoops 
#> Topic 5 Top Words:
#>       Highest Prob: you, i, the, to, a, what, it 
#>       FREX: will, mom, lucas, murderer, barb, he, know 
#>       Lift: demogorgon, shut, chug, missing, weird, probably, threeonefive 
#>       Score: threeonefive, mom, barb, lucas, murderer, hopper, sunflower

#_______
tidy(topic_model, matrix = "beta") %>%
  group_by(topic) %>%
  slice_max(beta, n = 10, ) %>%
  mutate(rank = row_number()) %>%
  ungroup() %>%
  select(-beta) %>%
  pivot_wider(
    names_from = "topic", 
    names_glue = "topic {.name}",
    values_from = term
  ) %>%
  select(-rank) %>%
  knitr::kable()
topic 1 topic 2 topic 3 topic 4 topic 5
you you you you you
i i i i i
the the the the the
to a to to to
a to a it a
and it and a what
it and that and it
that what it is and
what that is go that
it’s we we this is

#_________

tidy(topic_model, matrix = "frex") %>%
  group_by(topic) %>%
  slice_max(beta, n = 10, ) %>%
  mutate(rank = row_number()) %>%
  ungroup() %>%
  select(-beta) %>%
  pivot_wider(
    names_from = "topic", 
    names_glue = "topic {.name}",
    values_from = term
  ) %>%
  select(-rank) %>%
  knitr::kable()
#> Error in match.arg(matrix): 'arg' should be one of "beta", "gamma", "theta"

Created on 2022-12-18 with reprex v2.0.2

frex isnt available yet on the cran version of tidytext.

you can install the development version from GitHub with remotes:

library(remotes)
install_github("juliasilge/tidytext")

Well, there it is. Thanks nigrahamuk. I was scratching my head so much over this one that I made a hole.

Maybe there is still something more? After I ran the "install_github("juliasilge/tidytext")" command without error, and then reran the code, I now get this error at the same location in the code:

Error in slice_max():
! Problem while computing indices.
:information_source: The error occurred in group 1: topic = 1.
Caused by error:
! order_by must be a vector, not a function.

I'm afraid the hole will turn into a pit.

I guess that the slice_max fails because there is no beta column , so it thinks you are asking it to use the beta function. By analogy of the previous code, perhaps the column is now frex rather than beta

This topic was automatically closed 21 days after the last reply. New replies are no longer allowed.

If you have a query related to it or one of the replies, start a new topic and refer back with a link.