Major-Assignment 3
Yuhan Xu
2025-11-22
# Package names
packages <- c("RedditExtractoR", "anytime", "magrittr", "httr", "tidytext", "tidyverse", "igraph", "ggraph", "wordcloud2", "textdata", "here", "tidytext", "tidyverse", "textdata", "anytime", "magrittr", "wordcloud2", "ggdark", "syuzhet", "sentimentr", "lubridate", "here")
# Load packages
invisible(lapply(packages, library, character.only = TRUE))0. Describe in one sentence what you aim to examine using user-generated text data and sentiment analysis.
- Response: Sentiment towards PhD job market, focusing on academia.
- I did not specify a period because I am curious about the overall trend.
1. Search Reddit threads using a keyword of your choice.
- Specifying a subreddit for your search is optional.
- It is okay to combine data obtained by searching the keyword across multiple subreddits.
- You can choose any period, but ensure you gather a sufficient amount of data so that you can get meaningful results.
1-1. Downloading Reddit threads
# using keyword
threads_1 <- find_thread_urls(keywords = 'PhD job market',
sort_by = 'relevance',
period = 'all') %>%
drop_na()
rownames(threads_1) <- NULL
# Sanitize text
threads_1 %<>%
mutate(across(
where(is.character),
~ .x %>%
str_replace_all("\\|", "/") %>% # replace vertical bars
str_replace_all("\\n", " ") %>% # replace newlines
str_squish() # clean up extra spaces
))
colnames(threads_1)
head(threads_1, 3) %>% knitr::kable()# find_subreddits('PhD job market')
# search for subreddits
subreddit_list <- RedditExtractoR::find_subreddits('AskAcademia')
subreddit_list %>%
arrange(desc(subscribers)) %>%
.[1:25,c('subreddit','title','subscribers')] %>%
knitr::kable()threads_1$subreddit %>% table() %>% sort(decreasing = T) %>% head(20)# using subreddit
threads_2 <- find_thread_urls(subreddit = 'AskAcademia',
sort_by = 'top',
period = 'year') %>%
drop_na()
rownames(threads_2) <- NULL
# Sanitize text
threads_2 %<>%
mutate(across(
where(is.character),
~ .x %>%
str_replace_all("\\|", "/") %>%
str_replace_all("\\n", " ") %>%
str_squish()
))
head(threads_2, 3) %>% knitr::kable()# using both subreddit and keyword
threads_3 <- find_thread_urls(keywords= 'PhD job market',
subreddit = 'AskAcademia',
sort_by = 'relevance',
period = 'all') %>%
drop_na()
rownames(threads_3) <- NULL
# Sanitize text
threads_3 %<>%
mutate(across(
where(is.character),
~ .x %>%
str_replace_all("\\|", "/") %>%
str_replace_all("\\n", " ") %>%
str_squish()
))
head(threads_3, 3) %>% knitr::kable()Based on the collected threads, I decided to use
threads_3.
# write.csv(threads_3, "threads_3.csv", row.names = FALSE)
threads_3 <- read_csv("threads_3.csv")## Rows: 226 Columns: 7
## ── Column specification ────────────────────────────────────────────────────────
## Delimiter: ","
## chr (4): title, text, subreddit, url
## dbl (2): timestamp, comments
## date (1): date_utc
##
## 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.2. Clean your text data and then tokenize it.
2-1. Filter the text data without date
# create new column: date
threads_3 %<>%
mutate(date = as.POSIXct(date_utc)) %>%
filter(!is.na(date))2-2. Tokenization
# Word tokenization
words <- threads_3 %>%
unnest_tokens(output = word, input = text, token = "words")
# load list of stop words - from the tidytext package
data("stop_words")
# Regex that matches URL-type string
replace_reg <- "http[s]?://[A-Za-z\\d/\\.]+|&|<|>"
words_clean <- threads_3 %>%
# drop URLs
mutate(text = str_replace_all(text, replace_reg, "")) %>%
# Tokenization (word tokens)
unnest_tokens(word, text, token = "words") %>%
# drop stop words
anti_join(stop_words, by = "word") %>%
# drop non-alphabet-only strings
filter(str_detect(word, "[a-z]"))
# Check the number of rows after removal of the stop words. There should be fewer words now
print(
glue::glue("Before: {nrow(words)}, After: {nrow(words_clean)}")
)## Before: 58110, After: 19916I have removed the stop words, so I created a plot to see which meaningful words are used most frequently when people talk about PhD job market.
words_clean %>%
count(word, sort = TRUE) %>%
top_n(20, n) %>%
mutate(word = reorder(word, n)) %>%
ggplot(aes(x = word, y = n)) +
geom_col() +
xlab(NULL) +
coord_flip() +
labs(x = "words",
y = "counts",
title = "Unique wordcounts")
3. Generate a word cloud that illustrates the frequency of words excluding your keyword.
The word cloud highlights the dominant themes that appear in Reddit discussions about the PhD job market. The largest and most prominent words, such as research, career, time, tenure, positions, and field, indicate that users are primarily focused on employment pathways, the availability of academic positions, and long-term career planning.
remove_words <- c("phd", "job", "market", "academic", "academia")
words_clean %>%
filter(!(word %in% remove_words)) %>%
count(word, sort = TRUE) %>%
wordcloud2()4. Conduct a tri-gram analysis.
- Extract tri-grams from your text data.
- Remove tri-grams containing stop words or non-alphabetic terms.
- Present the frequency of tri-grams in a table.
- Discuss any noteworthy tri-grams you come across.
- If no meaningful tri-grams are found, you may analyze bi-grams as well. However, you still need to show results of the tri-grams.
# Get tri-grams.
words_ngram <- threads_3 %>%
mutate(text = str_replace_all(text, replace_reg, "")) %>%
select(text) %>%
unnest_tokens(output = paired_words,
input = text,
token = "ngrams",
n = 3)# Show tri-grams with sorted values
words_ngram %>%
count(paired_words, sort = TRUE) %>%
head(20) %>%
knitr::kable()| paired_words | n |
|---|---|
| the job market | 129 |
| i want to | 54 |
| i have a | 40 |
| job market is | 40 |
| on the job | 39 |
| a phd in | 38 |
| i don t | 37 |
| academic job market | 31 |
| the academic job | 30 |
| to get a | 30 |
| a lot of | 28 |
| in the us | 25 |
| job market in | 25 |
| i feel like | 23 |
| i know that | 23 |
| and i m | 22 |
| i am a | 22 |
| in my field | 22 |
| a tenure track | 20 |
| i m not | 20 |
#separate the paired words into three columns
words_ngram_pair <- words_ngram %>%
separate(paired_words, c("word1", "word2", "word3"), sep = " ")
# filter rows where there are stop words
words_ngram_pair_filtered <- words_ngram_pair %>%
# drop stop words
filter(!word1 %in% stop_words$word & !word2 %in% stop_words$word & !word3 %in% stop_words$word) %>%
# drop non-alphabet-only strings
filter(str_detect(word1, "[a-z]") & str_detect(word2, "[a-z]") & str_detect(word3, "[a-z]"))
# Filter out words that are not encoded in ASCII
# To see what's ASCII, google 'ASCII table'
library(stringi)
words_ngram_pair_filtered %<>%
filter(stri_enc_isascii(word1) & stri_enc_isascii(word2) & stri_enc_isascii(word3))
# Sort the new tri-gram counts:
words_counts <- words_ngram_pair_filtered %>%
count(word1, word2, word3) %>%
arrange(desc(n))
head(words_counts, 20) %>%
knitr::kable()| word1 | word2 | word3 | n |
|---|---|---|---|
| academic | job | market | 31 |
| tenure | track | position | 12 |
| tenure | track | job | 9 |
| current | job | market | 8 |
| assistant | professor | position | 4 |
| tenure | track | faculty | 4 |
| tenure | track | positions | 4 |
| tt | assistant | professor | 4 |
| english | adolescence | education | 3 |
| mental | health | struggles | 3 |
| peer | reviewed | journal | 3 |
| tenure | track | jobs | 3 |
| 20th | 21st | century | 2 |
| absolutely | love | teaching | 2 |
| academic | job | search | 2 |
| arts | humanities | discipline | 2 |
| author | conference | publications | 2 |
| business | specific | publications | 2 |
| changed | welsh | government | 2 |
| community | college | prof | 2 |
# plot word network
words_counts %>%
filter(n >= 3) %>%
graph_from_data_frame() %>% # convert to graph
ggraph(layout = "fr") +
geom_edge_link(aes(edge_alpha = .6, edge_width = n)) +
geom_node_point(color = "darkslategray4", size = 3) +
geom_node_text(aes(label = name), vjust = 1.8) +
labs(title = "Word Networks",
x = "", y = "")
Reddit discussions about the PhD job market, particularly within academic contexts, seem to center strongly on uncertainty. Conversations about securing a tenure track role appear repeatedly, with several high-frequency combinations such as tenure – track – position (12), tenure – track – job (9), tenure – track – faculty (4), tenure – track – positions (4), and tenure – track – jobs (3), suggesting ongoing concern about the scarcity of academic opportunities. Early-career stages also feature prominently, as indicated by assistant – professor – position (4) and tt – assistant – professor (4), which seem to highlight users’ focus on the transition into tenure-track faculty roles. Additional patterns like current – job – market (8) suggest that users may be comparing present conditions with prior expectations.
Some less frequent triads provide hints about the broader emotional and professional context. For example, mental – health – struggles (3) appears to point toward the psychological strain associated with job searching, and peer – reviewed – journal (3) may reflect concerns about publication expectations. Overall, the keyword frequencies suggest pervasive anxiety surrounding tenure track scarcity, the competitiveness of faculty hiring, and the stress embedded in discussions about the PhD job market.
5. Perform a sentiment analysis on your text data using a dictionary method that accommodates negations.
6. Display 10 sample texts alongside their sentiment scores and evaluate the credibility of the sentiment analysis outcomes.
- You are welcome to apply a deep learning-based model to enrich your analysis, but employing the dictionary method is imperative.
5/6-1. Using a dictionary method
# import the data
reddit_sentiment <- threads_3
reddit_sentiment %<>%
mutate(title = replace_na(title, ""),
text = replace_na(text, ""),
title_text = str_c(title, text, sep = ". "))
# dictionary method
reddit_sentiment_dictionary <- sentiment_by(reddit_sentiment$title_text)
reddit_sentiment$sentiment_dict <- reddit_sentiment_dictionary %>% pull(ave_sentiment)
reddit_sentiment$word_count <- reddit_sentiment_dictionary %>% pull(word_count)sentimentr_example <- reddit_sentiment %>%
mutate(sentimentr_abs = abs(sentiment_dict),
sentimentr_binary = case_when(sentiment_dict > 0 ~ 'positive',
TRUE ~ 'negative')) %>%
group_by(sentimentr_binary) %>%
arrange(desc(sentimentr_abs)) %>%
slice_head(n = 5) %>%
ungroup() %>%
arrange(sentiment_dict)
# negative
print("Negative sentiment")## [1] "Negative sentiment"sentimentr_example %>% filter(sentimentr_binary == 'negative') %>% pull(title_text) %>% print()## [1] "I m going to start my PhD. I graduated in chemical engineering with a masters degree at the end of 2023. I searched months for a bad paid job, in a field that has almost nothing to do with my field, is boring and paid bad. After a year i started applying for jobs again but the job market in germany is cooked. I applied for a phd position in the fusion reactor research (simulations). But i constantly worry that its not worth the risk and stress and at the end and waiting for a better job opportunity might just be better at least i got the sams pay i dont know, is there someone in a similar situation, are my worries based?"
## [2] "Are PhD programs considering reducing admissions amidst the impending collapse of higher education and the demographic crisis?. With the academic job market worsening, stagnant research funding, and a constant influx of PhD graduates for a limited number of faculty positions, the situation was already challenging. However, the COVID-19 pandemic has accelerated the bursting of the higher education bubble. Given these circumstances, are there any discussions in your departments about admitting fewer graduate students to prevent exacerbating the problem? Or is the allure of cheap graduate student labor too enticing to pass up? Additionally, do faculty members in your field discourage individuals from pursuing graduate school due to these challenges? Share your insights!"
## [3] "How to explain a poor relationship with an advisor in the job market?. Hi everyone, I will be graduating and entering the job market soon. One question I haven't resolved yet is how to explain the poor relationship I have with my PhD advisor. This isn't just a \"we didn't get along\" relationship, but a \"had to get the department chair and graduate school to bat for me\" kind of relationship. I probably won't be asking for a recommendation letter, so how can I explain this to interviewers? Is it alright to lay out the gory facts (succinctly) and let them decide?"
## [4] "Not even applying for academic jobs? Is it bad?. So I am finishing up my PhD in a social science field. Most of the people in my department try to find an academic job (at least they say so; it almost.feels like saying that you are not even looking for TT job seems to be politically incorrect). Anyways, so I never had interest in academic research at all ( the whole pubs cycle etc)., but recently after teachings 100s of entitled undergrads, I've lost my interest in teaching at all. So given that I don't like either teaching or research, I don't think I have a place in academia. Moreover, location is a big thing for me, I am a minority and don't want to live in a small college town forever. Anyways, so since I clearly know I am not feeling it for academia, I am not wasting my time applying for the jobs. However, my peers in my program, and even my advisor suggested me to apply for academic jobs, and only look for alt-acc if it doesn't work out. I've applied to a few, just in case, but I don't really want to waste my type applying for jobs that I don't even like the idea of working at, have no passion for, but the job market is so bad, and PhD makes us so overqualified, that I am looking for some academic jobs just to show my advisor. Overall, for those folks who wanted to not pursue academia, did you still apply for those jobs or you didn't even try?"
## [5] "Is the job market equally bad for all humanities PhDs, or are certain fields/areas/specializations more in-demand?. Some background: I have a bachelor's in history and bachelor's in religious studies. I'm currently working as an accountant, I hate my career, and I keep thinking about applying for a PhD in the humanities. I have been told by various people that a PhD in the humanities is a very bad idea since the job market is so brutal. But I've also been told by others that PhDs in certain fields (like Islamic studies, Arabic linguistics, Mandarin, or international relations with a focus on China or the Middle East) would be more valuable in the job market. Is this true? Would a PhD in Islamic studies for example be more valuable than a PhD in American history or English literature?"# positive
print("Positive sentiment")## [1] "Positive sentiment"sentimentr_example %>% filter(sentimentr_binary == 'positive') %>% pull(title_text) %>% print()## [1] "PhD advisor quitting academia when I am getting on the job market. Anything to plan ahead for?. Edit: Thank you so much for the kind and helpful responses. Now I have a better idea about what may happen next. My advisor and I will have further discussions in the next months. I am deleting the post for privacy concerns, but anyone in similar situations should still be able to refer to the suggestions in the comments. I am sad to see a brilliant researcher and mentor leaving the field, but it is probably better for them this way. Shit can happen in life, and self care is very important, whether you are in academia or industry."
## [2] "Received K99 but considering industry \024 worth activating or better to move on?. Hi all, I was recently awarded a K99/R00, and while I\031m really grateful for the support, I\031m now facing a tough decision: should I activate it or take a strong industry offer I\031ve received? I have a PhD in mechanical engineering, and my research has focused on acoustics, MEMS, and biomedical sensors. While I enjoy academic research and could see myself leading a lab someday, the current job market in engineering \024 especially outside of AI/data science \024 seems pretty discouraging. The industry role I\031m considering is exciting, well-paid, and offers more short-term stability. But I\031m struggling with the idea of walking away from the K99, especially given how competitive it is and how much effort I\031ve put into publications and grants over the last few years. So I\031m torn between: * Activating the K99 and trying again in the next faculty cycle (with the risk of landing in a holding pattern again), or * Moving on to a new challenge in industry. I\031m also curious if there\031s any precedent for a hybrid setup \024 like maintaining an academic affiliation while working in industry \024 and whether NIH or institutions are open to that. If anyone has been through something similar or has advice, I\031d really appreciate hearing your thoughts. Thanks!"
## [3] "As a PhD student, why is the recommendation to choose a \"well-known\" advisor so heavily emphasized?. Personal context is that I recently made a decision to attend a PhD program in which the university itself has much greater research/funding opportunities than the alternatives, but fewer choices of super \"high-impact\" advisors in my field. There were other reasons I made this decision, of course, but I ultimately went against the general guidance of \"your advisor matters more than the school\" and am second-guessing myself a bit. My question is, *why* is this the guidance exactly? I.E., what are the mechanisms in which this is a beneficial? To be clear, I do understand the importance of networking and strong LoRs. It also tracks that a well-known advisor will know a wider breadth of researchers to collaborate and publish with. But at the end of the day, in terms of landing a good placement in academia (or industry for that matter) will a \"famous\" advisor really carry that much more weight than a mid-career, respectably-published advisor? Or is it simply that a student will be more likely to garner peer-reviewed papers if they collaborate with a widely-cited senior author, and therefore, would be more competitive on the job market through their publications? I have heard quite a bit of conflicting advice regarding this dynamic, and really am just trying my best to navigate it all as I transition into my program. Any advice is very much appreciated!"
## [4] "Clinician (PA) Exploring a Pivot into Digital Health, Health Innovation & Global Opportunities \023 Seeking Program and Career Advice. Hi everyone, I\031m a U.S.-based Physician Assistant with a background in surgery and clinical care, currently exploring a major pivot into **digital health, global health, and health innovation**. I'm passionate about creating more sustainable, equitable health systems\024and I'm especially interested in how **technology, AI, policy, and implementation science** intersect to transform care delivery. I\031ve worked closely with robotics and surgical technologies in my clinical role, and I\031m now hoping to transition into a non-clinical career where I can work **internationally or remotely** at the intersection of **health equity, digital transformation, and innovation**. I\031ve been researching **MPH, MSc, and Health Informatics programs**, especially those with a focus on **digital health, AI for healthcare, and global health systems**. I\031m open to both U.S. and international programs\024including those in the **UK, Netherlands, Portugal, Ireland**, and beyond. That said, I\031ve seen a lot of mixed feedback on the value of MPH degrees in today\031s job market\024especially in the U.S.\024so I want to make sure I\031m being strategic in this pivot. **A few questions I\031d love insight on:** * If you\031ve made a similar pivot, what program (U.S. or international) helped you most\024and why? * Did your degree open doors into digital health, global roles, or AI-focused work? * Would you recommend other routes (certificates, fellowships, internships, short courses) instead of or in addition to a degree? * For those working in **global health, health tech, or remote roles**, how did you break in\024and do you have any advice (or regrets)? * Any specific programs or countries you found especially supportive or financially reasonable? I\031d really appreciate any thoughts, personal stories, or even program suggestions. Thanks in advance for sharing your experiences!"
## [5] "Questions About Job Prospects After A PhD. I\031m sure a question like this has been asked a million times before, but I\031m curious as to how this sub will respond in this specific case. So, my dream is to get a PhD studying Early Modern History, and I was wondering what the job prospects would be like after exiting grad school. I know the academic market is said to be extremely tough, but I wanted to get it directly from the source. For reference, I\031m currently a first-year studying at an Ivy League university, and hoping to attend a top grad school upon finishing undergrad. I understand that attending a prestigious university like this increases my chances of attending another prestigious university for grad school. I suppose I should be a bit more direct. Realistically, attending a top grad school, getting a good amount of research under my belt ( I know that\031s quite important), etc, what could I expect on the academic market?"I found that the dictionary-based sentiment analysis produced overly optimistic estimates that did not fully align with the actual tone of the discussions. To improve accuracy, I additionally applied a deep learning based sentiment analysis method, which provides a more context-aware interpretation of users’ expressions.
5/6-2. Import the result from Colab
# import the data
reddit_sentiment <- read_csv("threads_3_bert.csv")## New names:
## Rows: 226 Columns: 10
## ── Column specification
## ──────────────────────────────────────────────────────── Delimiter: "," chr
## (5): title, text, subreddit, url, bert_label dbl (4): ...1, timestamp,
## comments, bert_score date (1): date_utc
## 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.
## • `` -> `...1`# drop NAs
reddit_sentiment %<>% drop_na('bert_label')
reddit_sentiment %<>%
mutate(title = replace_na(title, ""),
text = replace_na(text, ""),
title_text = str_c(title, text, sep = ". "))5/6-3. Comparison with the dictionary method
# Join thread title and text.
reddit_sentiment %<>%
mutate(title = replace_na(title, ""),
text = replace_na(text, ""),
title_text = str_c(title, text, sep = ". "))
# dictionary method
reddit_sentiment_dictionary <- sentiment_by(reddit_sentiment$title_text)
reddit_sentiment$sentiment_dict <- reddit_sentiment_dictionary %>% pull(ave_sentiment)
reddit_sentiment$word_count <- reddit_sentiment_dictionary %>% pull(word_count)reddit_sentiment %<>% mutate(bert_label_numeric = str_sub(bert_label, 1, 1) %>% as.numeric())
bert_example <- reddit_sentiment %>%
filter(bert_label_numeric %in% c(1,5)) %>%
group_by(bert_label) %>%
arrange(desc(bert_score)) %>%
slice_head(n = 5) %>%
ungroup()
# 1 star
bert_example %>% filter(bert_label_numeric == 1) %>% pull(title_text) %>% print()## [1] "What is this cohort of graduating PhDs supposed to do?. This new wave of PhD students honestly feels cursed. They\031re like the pandemic\031s leftovers \024 high school during Trump, college during COVID, now graduating into yet another dumpster fire of an economy. Every step of their academic life has been some kind of hellscape. And yet universities keep cranking out PhDs like it\031s a factory line. It\031s insane. Every department is bloated with grad students, but the job market is a bloodbath. Tenure-track? Basically a lottery ticket. Industry? Doesn\031t want most of them. So what happens? Thousands of shiny new \034COVID-era PhDs\035 floating around with no real place to land, stuck in postdoc purgatory or adjunct hell until they burn out. At some point you\031ve gotta wonder: what\031s the endgame here? Because right now it looks less like \034training the next generation of scholars\035 and more like \034academic pyramid scheme with better branding.\035"
## [2] "What is the logic behind an unemployed PhD being \"stale goods on the market\"?. Quite a few times now I've heard people say that you need to land a position within a year or two of receiving your PhD, because otherwise you become \"stale goods on the market\" and... your PhD is... worth nothing, I guess?... Can someone explain to me how this makes any sense at all? It seems like the most utterly childish thing."
## [3] "Finished my PhD, now I'm just back to be unemployed.. Hey all, I am an EU citizen and I just finished a phd with an interdisciplinar research between music and environmental humanities in Ireland. Great feedback from peers, my research has been deemed very original and innovative, yet I feel incredibly stupid and incompetent now that I have to find a job. Postdoc research jobs are rare and applications are exhausting and ultra-competitive. I applied to several positions over the last few months, with no success. Exploring the job market outside of academia makes me feel really useless: none of the skills that I have are ever required by any employer, and seems that the only thing that I can do is go back to do hospitality jobs. Does anyone have any advice?"
## [4] "Funding running out. Not looking for advice\024just venting and checking in to see how everyone else is doing. I\031m in a soft money position at the medical school, and my K funding runs out this year. I submitted multiple NIH and private foundation grants last fall, but it\031s uncertain what will be reviewed, when, or if anything will be funded. I was supposed to have support through other projects, but everything is in flux. My work is in an area actively targeted by the new administration, and several collaborators have already had grant awards rescinded. I have a non-clinical PhD, so unlike some of my colleagues with MDs or clinical PhDs, cannot see patients to cover my salary. So, I guess I'm also back on the job market. I know I shouldn't complain too much because many of my non-academic colleagues working for gov't have been dealt a worse hand, but I just feel so tired. I've worked so hard, pushed out so many papers and grants, gone to the \"right\" places, done the right things, and here I am. It blows. I'm too old to have this much uncertainty in my life. How are you all faring? Hang in there."
## [5] "How do we know we're ready for a PhD?. Hey everyone, This is a super cliche and stupid question but it's something I don't know who to talk to. I'm close to finishing my 1st year of M.S. and want to get started early on making a decision for the future. I don't have internships or work exp. so that would make it challenging for me to secure a job in todays market (I'm in aerospace engineering) but I don't want to close that door. I am super confused if I want to do a PhD, on one hand I absolutely enjoy the deep technical work I have to do and honestly love it when my code produces the results I've been trying to get. But on the other hand, the same code and research has been burning me out. The code doesn't work like 8/10 times and I'm running out of those \"Aha!\" moments that take my research further. My research area is fairly unexplored so there isn't much I can read to make significant progress. Because of this I'm scared I might not be cut out for a PhD and it's making me question my abilities to do any research. If you're doing a PhD, did you face the same? How did you know if you were ready to commit 4-5 years? Thanks!"# 5 star
bert_example %>% filter(bert_label_numeric == 5) %>% pull(title_text) %>% print()## [1] "Tips for New Community College Professor. This past spring I landed a community college assistant professor (economics in a social science department) position which starts full time in a couple weeks! I have been teaching similar courses for the past several years at other colleges as an adjunct and PhD student, and I even was invited to start at my new place as an adjunct over the summer to teach a new class. While I know academia has its flaws, I've loved getting to work with students and help them grow their knowledge and apply economic thinking to their lives, moreso than being an economist myself. Also, in this job market, I cant even begin to discuss how grateful I am to have gotten this oppotunity. While I am very excited for my courses to start, and feel confident in tackling my teaching load and courses, I wanted to see if anyone would be willing to share tips to help me make the best of my early years or career! This position is actually something I have dreamed about and been working toward for a while now, so I would love to be the best professor I can be!"
## [2] "Job prospects for PhD in History/Philosophy of Science. I'm a STEM major (Organic Chemistry) who enjoys reading history of science. I'm not talking about strictly academic history, but books like A Short History of Nearly Everything, American Prometheus, Emperor of all Maladies, Einstein's Biography, etc. I looked up for fields which are at the intersection of science and humanities and History/Philosophy of Science came up. Some top schools have PhD programs in this discipline and I went through the coursework. It absolutely sounds fascinating! I follow this subreddit religiously and I hear the academic market for Historians (and in general, humanities) is horrendous. Which got me curious about the job prospects of a PhD in History/Philosophy of Science. I'm absolutely unaware about the field and what positions I could work in after completing it. What types of roles are available? Where can I find employment? And most importantly, how lucrative is it, because at the end of the fascination doesn't pay bills. Any advise, tips and insights would be extremely helpful. Thank you all! :)"
## [3] "Which Job Market?. Dear readers I need some information. I hope this vast world of a PhD community can help me, since my supervisor can not. I need to go to the job market in a bit more then a year. I am an interdisciplinary researcher in finance, economics and also have knowledge and experiments in the neuroeconomics. Most people in my department either go to industry or the ASSA. Which is the largest job market to go to. There are a few other job markets that I am aware of, like the European Economics one. However, I struggle to find any information on job markets other then the ASSA, which are related to business or managent or more finance oriented due to not knowing any names. Does anyone know any other large job markets? What are the names? I am asking this since a few people asked me if I was on the job market in October, while the only job market I was/am aware of is the ASSA in January. Can anyone help me? Much appreciated! :) Just the names suffice, I am already happy with any information."
## [4] "Last year of my Ph.D. in CS, stepping into the academia job market this fall. Last year Ph.D. in computer science at a university ranking 30-40 in U.S. Finally, I am stepping into the market. It matters a lot for me to give a try on the academia. I am preparing for my materials including CV, Research Statement, Teaching Statement, and so on. I saw many tutorials and personal experiences, which are really helpful. Do appreciate it if you could provide some advice on the research statement writing and mental adjustment during the job search process. Good luck with anyone who is searching as well!"
## [5] "Teaching at a Private College?. Current PhD candidate on the job market. I\031ve only attended public universities for my undergrad and grad programs. I applied to and now have an in person interview for an assistant professor position at a private liberal arts college. For those of you that teach in private schools, how different is it from a public university? Any information helps! Thank you!"7. Discuss intriguing insights derived from the sentiment analysis, supporting your observations with at least THREE plots.
7-1. Check the correlation between the sentiment values from two different methods.
reddit_sentiment %<>% mutate(bert_label_numeric = str_sub(bert_label, 1, 1) %>% as.numeric())
cor(reddit_sentiment$bert_label_numeric, reddit_sentiment$sentiment_dict)## [1] 0.3014132ggplot(data = reddit_sentiment, aes(x = bert_label_numeric, y = sentiment_dict)) +
geom_jitter(width = 0.1, height = 0) +
geom_line(aes(y = 0), color = '#FFD700', lwd = 1, linetype='dashed') +
dark_theme_grey()## Inverted geom defaults of fill and color/colour.
## To change them back, use invert_geom_defaults().
Similar to the lab’s results, my plot shows that the dictionary-based sentiment scores and the BERT sentiment labels do not seem to be correlated. The points are widely scattered across all five BERT label categories, with no clear upward or downward trend. Even within the same BERT class, the dictionary scores vary considerably, suggesting that the dictionary method does not capture the nuances that the deep learning model identifies.
7-2. Sentiment distribution
- Number of threads by sentiment category. The sentiment distribution is strongly skewed toward the negative end of the scale. The 2 star category is by far the largest group, with more than one hundred posts, suggesting that most discussions lean relatively negative rather than strongly negative or positive. The 4 star and 5 star posts are relatively rare, with the 5 star category being the smallest. Overall, the distribution suggests that conversations about the PhD job market on Reddit tend to express negative or somewhat negative sentiment, with limited evidence of positive emotional tone.
reddit_sentiment %>%
ggplot(aes(x = bert_label)) +
geom_bar(fill = "white") +
dark_theme_gray()
- Word counts by sentiment category. The figure suggests that word count does not vary in a strong or systematic way across sentiment categories. Posts in all five sentiment groups show wide dispersion, ranging from very short threads to posts exceeding three or four hundred words. The 2 star and 4 star posts appear to have slightly higher median word counts, while 3 star posts have a somewhat lower median. 5 star posts tend to be shorter on average, although the sample size for this category is small. Longer posts do not necessarily express more negative or more positive emotion, and users across sentiment categories seem similarly likely to write both brief and lengthy comments.
reddit_sentiment %>%
ggplot(aes(x = bert_label, y = word_count)) +
geom_jitter(height = 0, width = 0.05) +
stat_summary(fun = mean, geom = "crossbar", width = 0.4, color = "red") +
dark_theme_gray()
7-3. Temporal analysis
The two figures illustrate how discussions about the PhD job market on Reddit have evolved over time, both in terms of volume and sentiment composition. From the count plot, it is clear that the number of posts remained very low between 2012 and 2016, then began to rise gradually around 2018. This increase becomes much more pronounced in the most recent years, suggesting that public attention to the PhD job market has intensified only recently, likely reflecting worsening academic job prospects and broader post-pandemic uncertainties.
The proportional stacked bar plot further shows that sentiment toward the PhD job market is overwhelmingly negative across most years. The 1 star and 2 star categories consistently dominate, indicating that users primarily express worry, frustration, or pessimism. Only a few isolated years, such as 2013, 2016, and 2017, show a noticeable share of more positive sentiments (4 stars or 5 stars), and even these are limited and not part of a sustained trend. After 2019, negative sentiment becomes even more concentrated, with 2 star posts occupying the largest share year after year.
Overall, the figures suggest two key patterns. First, discussions about the PhD job market have increased substantially in the last five to six years. Second, the sentiment expressed in these discussions is predominantly negative, highlighting persistent anxiety and pessimism surrounding PhD employment prospects.
reddit_sentiment %<>%
mutate(date = as.POSIXct(date_utc)) %>%
filter(!is.na(date)) %>%
mutate(year = year(date))
# sentiment by year
reddit_sentiment %>%
filter(year >= 2012) %>%
ggplot(aes(x = year, fill = bert_label)) +
geom_bar(position = 'stack') +
scale_x_continuous(breaks = seq(min(reddit_sentiment$year),
max(reddit_sentiment$year),
by = 1)) +
scale_fill_brewer(palette = 'PuRd', direction = -1) +
dark_theme_grey() +
theme(
axis.text.x = element_text(size = 12, angle = 90, hjust = 1)
)
# sentiment by year
reddit_sentiment %>%
filter(year >= 2012) %>%
ggplot(aes(x = year, fill = bert_label)) +
geom_bar(position = 'fill') +
scale_x_continuous(breaks = seq(min(reddit_sentiment$year),
max(reddit_sentiment$year),
by = 1)) +
scale_fill_brewer(palette = 'PuRd', direction = -1) +
dark_theme_grey() +
theme(
axis.text.x = element_text(size = 12, angle = 90, hjust = 1)
)