What is Tang Ping? Investigating the Social Media Phenomenon Through Natural Language Processing

The phrase "Tang Ping" (躺平) has been gaining attention since 2023, particularly in the wake of China's stringent and protracted COVID-19 societal lockdown. Literally translated as "lying flat," Tang Ping has emerged as a popular sentiment among Chinese individuals on social media. Its significance is so profound that it now has its own Wikipedia entry (“Tang Ping,” 2023) and has even garnered the attention of the BBC (“China’s New ‘tang Ping’ Trend Aims to Highlight Pressures of Work Culture,” 2021).

As described in a BBC article, Tang Ping represents "an antidote to society's pressures to find jobs and perform well while working long shifts." It advocates for "not overworking, being content with more attainable achievements and allowing time to unwind." Wow, what a philosophical mindset.

While there's a fascinating story behind the term "Tang Ping," I'm not the best storyteller, and this isn't a research paper. So, if you're curious about China's history, social and economic background, COVID-related policies, or even the values and beliefs of Chinese youngsters, just type "Hey, please write an article about the cultural background of China" into any GPT chatbot. For now, let's dive straight into the research.

I think researchers are also intrigued by the Tang Ping phenomenon. Studies presented at the AASP2023 explored another word "Nei Juan" (内卷, involution, vicious competition) using historical data from the Google corpus (where the term's meaning has evolved over time), someone compared it with Tang Ping and seen Tang Ping as the opposite attitude. A recent study used survey experiment to look into public perceptions of Tang Ping, evaluate the concept as morally good or bad. There's another study that collected around 1000 user comments from Reddit and conducted a thematic analysis (Zhou, 2023). But what exactly is Tang Ping? To gain insights, we can examine how people have discussed it, particularly on Weibo, the Chinese social media platform where most of the Tang Ping "complaints" originated.

The Current Study

In this study, let's delve into the world of Tang Ping on a three-step journey:

1. I will collect recent Weibo posts about Tang Ping to understand how people are discussing it.

2. I will analyze the collected posts to identify the topics that are most closely associated with Tang Ping.

3. Finally, create a visual representation of the identified topics to see how they are related to each other.

Data

To gather fresh insights into this topic, I utilized open-source crawlers in Python and collected a total of 51,890 Weibo user posts (original, not reposts) spanning April to July 2023. As long as the post mentioned "Tang Ping" in Chinese, it was included. This period coincides with the surge in public concern surrounding China's rising unemployment rate, particularly among its young generations (~30%). If the Tang Ping concept aligns with our current perception, this dataset holds immense relevance for our analysis.

These posts were then cleaned (removing user information and punctuations) and combined into a single corpus, and tokenized using jieba (Sun, 2012/2023) in Python. Next, I trained the tokenized corpus using fastText model (Grave et al., 2018) in genism (Řehůřek & Sojka, 2010) using Continuous Bag of Words (CBOW) in 200 dimensions, with a minimal word frequency of 10, 5 epochs, a window of size 5 and 10 negatives. The resulting output was a word embedding model, where each word token was represented by a vector of numerical values (I will just call it "model").

Analyses and Results

First, let's examine the top 100 words that exhibit the highest cosine similarity with the term "Tang Ping". You can download the whole list here: https://doi.org/10.17605/OSF.IO/HG2EP

The top 10 words already provide interesting glimpses into the core concepts that underpin the Tang Ping phenomenon:

As you can see, the first 10 words already give us clues about the amazing ideas behind Tang Ping. Most of the words are about being lazy and comfortable. The third word, "Tang," is also part of the phrase "Tang Ping," and sometimes people use short forms of it as slang. It might be hard to say exactly what it means, because Chinese words can have a lot of different meanings (okay...fewer words = richer meaning). But if you just think of it as the act of "lying down," you can already imagine what these 10 words mean together.

In fact, let's put the 10 words back to DALL.E and simply ask for a picture.

A picture is worth a thousand words.

In our list, there were some emojis, which I didn't mean to leave out, as mentioned earlier.

The other 90 words were also fascinating. They didn't just describe actions but also personality traits. These words perfectly fit into what I call 'The Chinese Tang Ping Scale'. Some examples include words like 睡懒觉 (sleeping in), 放弃 (giving up), 无所事事 (doing nothing), and 咸鱼 ("salted fish", failure), which suggest a sense of resignation or inactivity.

*I'm surprised that someone had already done this, a scale of Tang Ping! https://www.mdpi.com/2076-328X/13/11/915

Others such as 一动不动,安逸,破罐子破摔,吃喝玩乐,无忧无虑,算了吧,混日子,开开心心,做梦,放平,高不成低不就,下辈子,废物,不思进取,不求上进,无欲无求,颓废. All right, too much negative energy here.

For a different angle, let's consider the top 100 words related to the single character "Tang" from the same dataset. You can access the list here:[https://doi.org/10.17605/OSF.IO/HG2EP]. Most of these words are simple, focusing on actions like lying down and sleeping, with few touching on deeper personality or values. This observation aligns with our initial thoughts.

Next, we'll explore the top 100 words associated with Tang Ping, including Tang Ping itself, making it 101 words. Each word is represented in a 200-dimensional space. We'll then use T-SNE to reduce this to a 2-dimensional view, making it easier to understand. The results of this analysis are presented below:

These clusters were arbitrary but informative, showing groups of topics nested within the grand topic of Tang Ping.  Observing those big clusters, it seems that they are telling stories about "working class people", "a break from work", and "young people just starting their jobs". Well, the interpretation is yours.

Let's do some more comparisons. We can pick some words from the list, as well as some others words that are not "aligned" with the Tang Ping value, and see how they are related. Of course, this is not a comprehensive corpus with only 21,049 tokens, we cannot test everything. For simplicity, I will just go with the top 10 words, and some words about "positive energy", such as 吃苦 ( endure hardship), 吃苦耐劳 ( hardworking and enduring), 勤奋 (diligent), 迎难而上 (face difficulties head-on), and perhaps scenarios like 结婚 (get married), 工作 (work), 读书 (study). The vector values of these words were extracted and submitted to network analysis.

Within these networks, the nodes and edges provide insights into the level of associations between concepts. Nodes that exhibit stronger connections and relevance are positioned closer together, while those with weaker associations are dispersed. The thickness of the edges reflects the strength of the partial correlation between nodes. Notably, red edges represent negative correlations, whereas blue edges signify positive correlations.

So, it's pretty clear that Tang Ping is a whole different vibe compared to my hand-picked words, especially when the network graph put them side by side. And it seems like marriage and Tang Ping don't exactly mix – not that surprising, right? The whole positive energy cloud seems to steer clear of Tang Ping, and the angels and demons are neatly divided by 'Work' and 'Schooling (reading books).

Conclusions

My preliminary analysis of the Chinese term "Tang Ping," a rising concept in China that has sparked much discussion, appears to align with our initial observations and expectations. This bottom-up empirical approach has provided a rich and detailed understanding of the concept, allowing us to visualize the scenes and nuances associated with Tang Ping.With all these details, we can start the "I" (not AI) in our brain to grasp the essence of the Tang Ping spirit. This approach, I believe, offers a more objective and engaging perspective on this social phenomenon compared to traditional survey methods or interview studies.

These words, like"lay down," "slack off," "don't strive," "don't make a fuss," and "low desires," paint a vivid picture of the Tang Ping mindset – one that pictures the completely giving up, or prioritizes inner peace and detachment from external pressures. What do you think?

If you find this study to be of interest and would like to cite it, you can simply use the provided DOI: https://doi.org/10.17605/OSF.IO/HG2EP

High-resolution model results are available at the DOI provided above. While I've removed user IDs to protect privacy, I won't be sharing the raw data (user posts) publicly due to the potential presence of identifying information (yes, still! I found some posts mentioned names...). However, if you're interested in exploring the findings and testing the model, the pre-trained fastText model is available for download.

References

China’s new “tang ping” trend aims to highlight pressures of work culture. (2021, June 3). BBC News. https://www.bbc.com/news/world-asia-china-57348406

Řehůřek, R. (2022). Gensim: Topic modelling for humans. https://radimrehurek.com/gensim/auto_examples/tutorials/run_fasttext.html

Grave, E., Bojanowski, P., Gupta, P., Joulin, A., & Mikolov, T. (2018). Learning Word Vectors for 157 Languages. Proceedings of the International Conference on Language Resources and Evaluation (LREC 2018). https://fasttext.cc/docs/en/crawl-vectors.html

Sun, A. (2023). Jieba [Python]. https://github.com/fxsjy/jieba (Original work published 2012)

Tang ping. (2023). In Wikipedia. https://en.wikipedia.org/w/index.php?title=Tang_ping

Zhou, Y. R. (2023). The lying flat movement, global youth, and globality: A case of collective reading on Reddit. Globalizations, 20(4), 679–695. https://doi.org/10.1080/14747731.2023.2165377