Sentiment
Analysis

Day 15

Prof Amanda Luby

Carleton College
Stat 220 - Spring 2025

Today

Sentiment analysis
Lab Quiz 2!

Data for today

Random (?) sample of 26,882 reviews of coursera courses

en_coursera_reviews <- read_csv("https://stat220-s25.github.io/data/en_coursera_sample.csv")
en_coursera_reviews

# A tibble: 26,882 × 5
   CourseId                    Review                      Label cld2  review_id
   <chr>                       <chr>                       <dbl> <chr>     <dbl>
 1 nurture-market-strategies   It would be better if the …     1 en            1
 2 nand2tetris2                Superb course. Great prese…     5 en            2
 3 schedule-projects           Excellent course!               5 en            3
 4 teaching-english-capstone-2 I'd recommend this course …     5 en            4
 5 machine-learning            This course was so effecti…     5 en            5
 6 python-network-data         Words cannot describe how …     5 en            6
 7 clinical-trials             Great course!                   5 en            7
 8 python-genomics             I didn't know anything abo…     3 en            8
 9 strategic-management        Loved everything about thi…     5 en            9
10 script-writing              No significant instruction…     1 en           10
# ℹ 26,872 more rows

Are these positive or negative reviews?

“thank you so much it was great course”

“Too reliant on materials directly collected by professor, too little context of larger artistic community.”

“Excellent! easy to follow and engaging!”

“Slow and redundant. Would have preferred a faster-paced and more substantive course.”

“Way too shallow and way too english :(”

Sentiment analysis

One way to analyze the sentiment is to consider the text as a combination of the individual words. It’s computationally convenient because we’re able to use the tidy tools that we’ve been building up:

tokenize the text
join the sentiment values to each token
group the words in the document to summarize the overall sentiment

Sentiment datasets: `bing`

library(tidytext)
get_sentiments("bing") %>%
  slice_sample(n = 20)

# A tibble: 20 × 2
   word          sentiment
   <chr>         <chr>    
 1 undercutting  negative 
 2 unexpectedly  negative 
 3 slack         negative 
 4 fabrication   negative 
 5 majesty       positive 
 6 perilously    negative 
 7 cave          negative 
 8 partisan      negative 
 9 prik          negative 
10 miserableness negative 
11 flourish      positive 
12 acridness     negative 
13 darling       positive 
14 undisputed    positive 
15 crabby        negative 
16 inexperience  negative 
17 boastful      negative 
18 unequivocally positive 
19 greatness     positive 
20 gratify       positive

Sentiment datasets: `afinn`

library(textdata)
get_sentiments("afinn") %>%
  slice_sample(n = 20)

# A tibble: 20 × 2
   word           value
   <chr>          <dbl>
 1 postponed         -1
 2 beautifully        3
 3 chagrin           -2
 4 appreciation       2
 5 spark              1
 6 mope              -1
 7 empathetic         2
 8 oversimplified    -2
 9 stimulating        2
10 exaggerates       -2
11 disconsolation    -2
12 sophisticated      2
13 fearing           -2
14 bereaving         -2
15 dejected          -2
16 stopped           -1
17 strikers          -2
18 scary             -2
19 worse             -3
20 protesters        -2

en_coursera_reviews

# A tibble: 26,882 × 5
   CourseId                    Review                      Label cld2  review_id
   <chr>                       <chr>                       <dbl> <chr>     <dbl>
 1 nurture-market-strategies   It would be better if the …     1 en            1
 2 nand2tetris2                Superb course. Great prese…     5 en            2
 3 schedule-projects           Excellent course!               5 en            3
 4 teaching-english-capstone-2 I'd recommend this course …     5 en            4
 5 machine-learning            This course was so effecti…     5 en            5
 6 python-network-data         Words cannot describe how …     5 en            6
 7 clinical-trials             Great course!                   5 en            7
 8 python-genomics             I didn't know anything abo…     3 en            8
 9 strategic-management        Loved everything about thi…     5 en            9
10 script-writing              No significant instruction…     1 en           10
# ℹ 26,872 more rows

`inner_join`

`join`ing the sentiment data

bing_sentiments = get_sentiments("bing")
en_coursera_reviews %>%
  unnest_tokens(word, Review) %>% 
  inner_join(bing_sentiments, by = "word") %>%
  select(-c(CourseId, cld2))

# A tibble: 69,052 × 4
   Label review_id word        sentiment
   <dbl>     <dbl> <chr>       <chr>    
 1     1         1 better      positive 
 2     5         2 superb      positive 
 3     5         2 great       positive 
 4     5         2 challenging negative 
 5     5         2 fun         positive 
 6     5         2 recommend   positive 
 7     5         3 excellent   positive 
 8     5         4 recommend   positive 
 9     5         4 enjoyed     positive 
10     5         4 great       positive 
# ℹ 69,042 more rows

`group_by` review ID, `summarize` overall sentiment

bing_review_scores <- en_coursera_reviews %>%
  unnest_tokens(word, Review) %>% 
  inner_join(bing_sentiments, by = "word") %>%
  group_by(review_id) %>%
  summarize(
    sum = (sum(sentiment == "positive") - sum(sentiment == "negative"))
  )

bing_review_scores

# A tibble: 25,384 × 2
   review_id   sum
       <dbl> <int>
 1         1     1
 2         2     3
 3         3     1
 4         4     3
 5         5     5
 6         6     1
 7         7     1
 8         8     2
 9         9     5
10        10     0
# ℹ 25,374 more rows

Most positive review (by `sum`):

en_coursera_reviews %>%
  left_join(bing_review_scores, by = "review_id") %>%
  slice_max(sum) %>%
  pull(Review)

[1] "This course is awesome on so many levels. This is the best inferential statistics course I've come across. Here's why:*** The slides are beautiful and visually appealing, making following the rigorous content easier to digest.*** Instructors are captivating and articulate, the explanations are clear and concise.*** The assignments are very very tough, making the course incredibly challenging, but worth it. Honestly, I don't get why people give 1 star because the course is tough. This should be a huge plus.It was a real challenge getting 100% for everything. For every quiz, I attempted 2 - 3 times to get 100%. The challenge is worth it. I couldn't thank you enough for this course. You explain tough statistical concepts like the difference between prediction intervals and confidence intervals really well. Also, I think this course has the best teaching for Analysis of Variance (I have taken a few other statistics moocs). Also, your course helped me appreciate the meaning of R-squared, standard errors, confidence intervals in a very intuitive fashion. There are many other new things I've learnt from your course, some of them I thought I knew, but you helped me to either \"Aha\" or understand them more deeply.Before this course, most of the time statistics to me is like plug-and-play using procedures and and softwares. But now, I can understand the concepts and what the calculations really mean.Thank you for creating quizzes that make us really do step-by-step calculations and not just plug data into equations to get results like so many other statistics moocs do.The pedagogy is really great. Sometimes quizzes can be frustrating because I need to read very carefully into the meaning of the questions and all the options. However, the learning experience is really worth it.Again, thank you for an amazing course! This is rare stuff!It is without a doubt, a lot of passion and effort has been put into this course and this series."

Most negative reviews (by `sum`):

en_coursera_reviews %>%
  left_join(bing_review_scores, by = "review_id") %>%
  slice_min(sum) %>%
  pull(Review) %>%
  pluck(1)

[1] "Complex concepts (e.g., regression), which cannot be taught in such a short course format (I know from extensive prior training), are taught here in an arcane way. I don't know how students without prior knowledge on these topics (e.g., regression, ROC curve analysis) could possibly understand this when taught this way.Course is also in an 'early draft' mode, with plenty of mistakes in the videos/slides. The course (and specialization) really tarnishes - instead of enhancing - this institution's outside image and reputation.Really a deception; sorry I took this class and specialization, a complete waste of my time and money. The initial presentation was misleading (and I have found online many people sharing the same feeling)."

What weaknesses do you notice about this analysis?

What could we change?

Sentiment Analysis

Today

Data for today

Are these positive or negative reviews?

Sentiment analysis

Sentiment datasets: bing

Sentiment datasets: afinn

inner_join

joining the sentiment data

group_by review ID, summarize overall sentiment

Most positive review (by sum):

Most negative reviews (by sum):