“Go eat a bat, Chang!”: On the Emergence of Sinophobic Behavior on Web Communities in the Face of CO

Link: https://dl.acm.org/doi/10.1145/3442381.3450024

Conference: WWW 2021

Keywords: COVID-19, Sinophobia, Hate Speech, Twitter, 4chan

My notes

Summary

Collect 5 month post data from 2 online communities (4chan /pol/ and Twitter), measure the trends of Sinophobic, and investigate new terms with word2vec.

Related Technique: crawler/scrapper, word2vec, similarity calculation, data visualization

Pros

1. Hot topic from special perspectives (hatespeech and conspiracy theories) target to sinophobic

2. Nice intuitive visualization

3. Comperehensive investigation with NLP technique (word2vec)

Cons

1. Some statements are subjective inferences without groundtruths. Some slurs target to Asians insted of Chinese only (e.g. zipperhead, ricenigger)

2. Didn't explain pros of techniques they adopted

Details

Abstract

Target to the spread of COVID-19, to keep social distance, online medium became more active, but with spreading potentially harmful and disturbing content (e.g. conspiracy theories and hateful speech).

We collected and analyzed 2 larrge datasets from Twitter and 4chan (time period: 5 months) to investigate whether there is a rise or important differences with regard to the dissemination of Sinophobic content.

As a conclusion, COVID-19 boosted Sinophobia online and it is a cross-platform pheomenon (from fringe web communities like /pol/ ( "Politically Incorrect") to mainstream platforms like Twitter).

Findings

1. Discussions related to China and Chinese people on Twitter and 4chan's /pol/ after the outbreak of the COVID-19 pandemic raised. The increase in these discussions and Sinophobic slurs coincides with real-world events related to the COVID-19 pandemic.

2. With word embeddings, we find that various racial slurs are used in these contexts on both Twitter and /pol/ which shows it is a cross-platform phenomenon.

3. Analyzing word embeddings over time, we discover new emerging slurs and terms related to Sinophobic behavior, as well as the COVID-19 pandemic (e.g. asshoe, Kungflu).

4. By comparing data before and after COVID-19 outbreak, we observe shifts in the content posted by users on Twitter (towards blaming China and Chinese people) and /pol/ (towards using more and new Sinophobic slurs.).

Data

1. Twitter: Collection Tool: Streaming API Time Period: 1 Nov 2019 - 22 Mar 2020 Amount: 222,212,841 tweets (English only)

2. 4chan /pol/: Collection Tool: JsonAPI Time Period: 1 Nov 2019 - 22 Mar 2020 Amount: 16,808,191 posts.

Temporal Analysis

1. The rises on 4chan match the real-world major events Mentions of the terms “china” and “chinese” on 4chan's /pol/

Mentions of the terms “china” and “chinese” on 4chan's /pol/.

Major events

#	Date	Event
1	2019-12-12	President Donald Trump signs an initial trade deal with China.
2	2020-01-23	The Chinese government announces a lock-down in Wuhan.
3	2020-01-30	The World Health Organization declares a public health emergency].
4	2020-02-23	11 municipalities in Lombardy, Italy are locked down [19].
5	2020-03-09	Italy extends restrictions in the northern region of the country.
6	2020-03-16	Donald Trump referred to COVID-19 as “Chinese Virus” on Twitter.

2. Twitter discussion during the first peak is significantly lower than when Europe gets involved, possibly due to Twitter's broader geographic reach or /pol/'s susceptibility to conspiracy and racism-related posts on 4chan. Mentions of the terms “china” and “chinese” on Twitter.

3. Temporal dynamics of Sinophobic racial slurs on 4chan's /pol/ and Twitter

Slur pick: "chink,” “bugland,” “chankoro,” “chinazi,” “gook,” “insectoid,” “bugmen,” and “chingchong”

Mentions of Sinophobic racial slurs on 4chan's /pol/.

Mentions of Sinophobic racial slurs on Twitter.

Top 20 most similar words, along with their cosine similarities, to the words “china,” “chinese,” and “virus” obtained from the word2vec models trained for the whole period (November 2019 - March 2020).

Top 20 most similar words, along with their cosine similarities, to the words “china,” “chinese,” and “virus” obtained from the word2vec models trained for the whole period (November 2019 - March 2020).

Takeaway of Temporal Analysis

1. 4chan and Twitter are heavily discussing China in relation to COVID-19, and that this discussion accelerated rapidly once the Western world became affected.

2. Differences in the use of slurs on /pol/ and Twitter.

Content Analysis

Technique: word2vec (skip-gram) Method: 3 groups of word2vec models for each of Twitter and /pol/:   Group1: $W_a$ trained on all posts made during the period between October 28, 2019 and March 22, 2020 to study the use of words for the entire duration of our study.   Group2: One distinct word2vec model for each week between October 28, 2019 and March 22, 2020, $W_{t=i}$, i∈T (i is the ith week in T). To study changes in the use of words over time.   Group3: $W_c$ trained on all posts (July 1, 2016 to November 1, 2019) to investigate the emergence of new terms during the period of our study

Result: Visualization of a 2-hop graph from the word “chinese” on 4chan's /pol/ and Twitter.

2 categories of profanities:

1. Insults addressing Asian people: Racist variations of “china” and “chinese" (e.g. chingchong) Culturally oriented racist terms (e.g. ricenigger, yellownigger)

2. Sexual stereotypes (e.g., pindick)

Content Evolution

4chan pol

Top 20 most similar words, along with their cosine similarities, to the words “china,” “chinese,” and “virus” for the first and last trained word2vec models from 4chan's /pol/.

Visualization of a 2-hop graph from the word “chinese” on 4chan's /pol/ using the first and last weekly word2vec models.

1. 4chan's /pol/ users typically use racial slurs targeted to Chinese people even before the outbreak of the COVID-19 pandemic.

2. A rise in the use of this term in discussions related to China.

3. Slurs used against Chinese people increased.

New Term Discovery Compare $W_{t=i}$ and $W_c$, filter out cosine similarity below 0.5's terms. Visualization of the emergence of new words related to “chinese” over time on 4chan's /pol/.

Twitter

Top 20 most similar words, along with their cosine similarities, to the words “china,” “chinese,” and “virus” for the first and last trained word2vec models on Twitter. Visualization of the emergence of new words related to “chinese” over time on Twitter.

New Term Discovery During important real-world events, such as the COVID-19 pandemic, language evolves and new terms emerge on Web communities like Twitter.

Semantic Changes between Words

For each $W_{t=i}$, we extract the cosine similarity between two terms and then we plot their similarities over time. Finding: With time going, more similarity between "Chinese" and "Virus","Chink"

Cosine similarities between various terms over time.

Cosine similarities between various terms over time.

Conclusion

1. The paper aims to understand Sinophobic language related to COVID-19 on the social web.

2. The study collects data from 4chan's /pol/ and Twitter over five months, with word embedding, revealing a rise in Sinophobic content which is a crossplatform phenomenon on fringe web communities and mainstream platforms.

3. Sinophobic behavior evolves quickly and substantially, especially after world changing events like the COVID-19 pandemic.