How Different Genders Use Profanity on Twitter?

Abstract

Social media, is often the go-to place where people discuss their opinions and share their feelings. As some platforms provide more anonymity than others, users have taken advantage of that privilege, by sitting behind the screen, the use of profanity has been able to create a toxic environment. Although not all profanities are used to offend people, it is undeniable that the anonymity has allowed social media users to express themselves more freely, increasing the likelihood of swearing. In this study, the use of profanity by different gender classes is compiled, and the findings showed that different genders often employ swear words from different hate categories, e.g. males tend to use more terms from the "disability" hate group. Classification models have been developed to predict the gender of tweet authors, and results showed that profanity could be used to uncover the gender of anonymous users. This shows the possibility that profiling of cyberbullies can be done from the aspect of gender based on profanity usage.

Analysis of Popular Social Media Topics Regarding Plastic Pollution

“…The number of tweets provided for behavioural research studies commonly ranges from 50,000 to 200,000 tweets [63][64][65]. In this study, 274,404 tweets were crawled from Twitter using keywords between 16 April and 4 June 2020.…”

Scoping Review of Methods and Annotated Datasets Used to Predict Gender and Age of Twitter Users

“…3.2 Studies developing methods for gender and age prediction 3.2.1 Gender 44 studies developed ad-hoc methods to predict the Twitter user's gender. Of these, 32 predicted only gender 28,29,31,33,36,37,47,48,50,51,54,55,58,60,61,64,65,68,71,72,75,81,[83][84][85][86]90,92,94,96,100,101 and gender was predicted along with the user's age in 12 34,44,49,59,62,66,80,80,87,89,91,95 .…”

“…For the training and validation of the approaches for gender detection, some studies used previously created annotated corpora, while others collected data directly from Twitter. Among the 19 studies that used previously annotated data sets, nine 34,36,37,47,49,65,66,75,100 used corpora from the PAN-CLEF author profiling tasks [102][103][104][105][106][107][108] , while ten studies 51,54,62,64,72,83,84,94,96,101 relied on data sets from other studies 92,[109][110][111][112][113][114][115] .…”

“…Despite data limitations from the Twitter API, it was the main source of data collection, with 18 studies 28,30,31,33,48,50,51,55,58,62,68,71,[85][86][87]95,96,100 collecting data either as a random sample from the Twitter Streaming API or based on keywords or geographic location from the Twitter Search API. One study 61 collected data using a scraping tool, three 59,91,92 used a random sample from a collection of 10% of tweets from 2014-2017 or the Twitter archive, and one did not specify its data source 44 .…”

“…There were 11 studies that automatically, or (semi)-automatically, labeled their data sets through the detection of selfreports or gender identifying terms (e.g., mother, son, uncle, etc.) 48,59,87,89,91,96 , the user's name 28,86,92 or declarations on other linked social media 95,96 . While three studies created their labeled datasets by using the accounts of famous social media influencers 57 or other unspecified collection of users whose gender is known 28,43 .…”