“ The Four Letter Word ”

If you are an incessant user of social media then, odds are high that Virat Kohli popped into your feed. He had to face a backlash for his recent Diwali advertisement. It didn’t just stop there, a few maniacs on social media platforms trolled his daughter. Similarly, Fabindia had to deal with controversy regarding the name of their new line of clothes that they launched during the Diwali season. In an attempt to make this collection seem chicer, Fabindia named it “jashn-e-riwaaz”. It doesn’t require a genius to know how it triggered a controversy in the Indian community. You see, in both, these cases hate speech, usage of slurs (“Profanity” so to say) on social media are prominent. Lately, our big social media platforms are becoming more of a dystopia where backlashes, cyberbullying, and sharing of objectionable or inappropriate content are predominant. Online interactive environments have become an integral part of our lives, and to let such dreadful content prevail over these online networks and collaborative spaces is languishing. Coming across vexatious stuff very often affects different masses and different people of different age groups in manifolds.

What is the definition of Hate speech?

Definition of Hate speech becomes subjective and has different meanings and relevance to everyone. The demarcation between hate speech and free expression is not precise enough. Moreover, no institute provides a vivid definition of Hate speech. Vaguely put it is a direct attack or a violent threat or an insult at someone on the basis of attributes such as race, religion, age, gender, disabled/ diseased, geographic origin, etc. Dehumanizing and making someone feel inferior are also considered hate speech.

A research article published by individuals affiliated to Information Retrieval Laboratory, Georgetown University, Washington, DC, United States of America summarizes the complications that come across which defining hate speech in the following paragraphs

A difficulty that isn’t addressed by many definitions is that of factual claims. “Jews are swine,” for example, is plainly hate speech by most definitions (it is a derogatory statement), but “Many Jews are lawyers” is not. In the latter instance, we would need to evaluate whether each remark is factual or not utilising external sources to establish whether it constitutes hate speech. This form of hate speech is challenging because it involves real-world fact verification, which is itself a difficult process. Furthermore, in order to assess validity, we would need to first specify exact word interpretations, such as whether “many” refers to an absolute number or a percentage of the population, further complicating the task. Furthermore, in order to assess validity, we would need to first specify exact word interpretations, such as whether “many” refers to an absolute number or a proportion of the population, further complicating the verification.

Another problem with the concept of hate speech is the potential for glorifying a nasty group. Applauding the KKK, for example, is hate speech; nevertheless, praising another group is clearly non-hate speech. In this scenario, it’s crucial to understand which groups are hate groups and what exactly is being appreciated about them, as some of the praise is undeniably, and tragically, genuine. In terms of their “Final Solution,” the Nazis, for example, were quite efficient. As a result, processing praise on its own can be challenging at times.

The posed Dilemma…. Is profanity fair?

Studies have shown that swearing builds our tolerance level to physical pain, it builds our moral and emotional strength. Swearing can also be practiced as a way of calming ourselves down when in situations when we feel helpless.

Research at Keele University revealed that swearing provides short-term pain relief. it helps us deal with our high-pressure situations and overwhelmed emotions. This research article further talks about the possible reasons for the pain reliever effect under high pressure and muscular activities.

Increased muscular performance could also be attributed to a widespread disinhibition caused by swearing. While most people think of disinhibition in terms of psychological disinhibition, in which one’s inner self-control is less paid to, swearing may cause more general disinhibition, in which somatic self-monitoring is less attended to.

Swearing has a distinct tone and articulation that is not found in non-swear terms, for example, plosiveness (i.e. a speaking sound made by completely closing the oral channel and then releasing it with a burst of air). While many non-swear words are also plosive, a comprehensive assessment of plosiveness would be an intriguing area of investigation.

Finally, this research includes two tests that show that swearing can improve physical performance depending on muscular force. Participants were able to generate more force on a high-resistance bike pedalling activity and a firmer hand grip when repeating a curse phrase when compared to a non-swear word. However, improved physical performance occurred in the absence of observable changes in cardiovascular or autonomic activity, implying that the basic mechanism behind the effect of swearing on physical performance may be something other than sympathetic activation.

But the question stays intact: is it fair to let this short-term pain reliever catalyze hate speech?

Hate Speech and how it triggered horrific events

Our ideology mostly depends on what we see, what we hear, what we read, and what we feel. Now when much of the world communicate on the internet, what we see are viral videos, what we hear are controversies, what we read are the posts made by psychopaths from around the world. Finally what we feel is anxiety and depression (if you don’t feel it then, kudos to you buddy for making it into the tiny fraction of happy mankind).

Ultimately our motives are developed in an undesirable direction. Here are a few issues that were fueled by hate speech-

• An investigation made by a select committee on Jan.6 capitol attack revealed the companies like twitch, parler, Facebook, Twitter, 4chan, TikTok with their social media platforms played a critical role in spreading hate speech and misinformation.
• A recent paper leak by the FB whistleblower cited how RSS-linked pages in India have been promoting an “Anti-Muslim narrative” and creating a drift between the religious communities.
• In Myanmar, platforms like Facebook are playing a central role in amplifying the already existing rifts between the Buddhist majority and minority Rohingya Muslims.
• In 2019 New Zealand mosque shooting was broadcast on youtube.

These are just a few incidents, similar ones took place in Germany, Sri Lanka, France, etc.

Such infuriating content should be discouraged and a friendly environment promoting peaceful information should be built. And recognizing such depreciated content and managing is a challenge. This calls for technologies like Profanity filter and content moderations.

Hate speech detection -

The research article which we talked about in the definition section also provided insights on how hate speech can be detected. They have broadly put the automatic hate speech detection approaches into 3 categories: Key-based approach, source metadata, and machine learning classifiers.

First key-based approach: It solely relies on a blacklist of words or phrases that are potentially hateful or offensive and creates an alert and hides them accordingly. This approach has a pretty simple algorithm and is fast too but it has its own set of complications. They rely one hundred percent on the list and can’t interpret the context. For instance the word “bitch” is a slur but it doesn’t carry an offensive meaning when used for a dog. Moreover, languages are evolving, people come up with new slang and creative misspellings or they use the symbols in a clever manner.

Second source metadata: This is rather a more dynamic approach that takes into account the demographics, timestamp, social engagement, or locations. Simply put, a user with a track record of spreading hate speech and a user with no such record will be treated in a different manner while blocking their content even if their posts are ditto. This system might seem flawed as it creates a bias across different groups of people. It ultimately puts borders on free expression.

Lastly Machine Learning Classifiers: There are numerous ML algorithms available that can be deployed for automated categorization of hateful text. Research paper — Automatic Hate Speech Detection using Machine Learning: A Comparative Study in (IJACSA) International Journal of Advanced Computer Science and Applications, cited SVM and RF algorithms have a better performance over others like KNN, NB, LR, DT, Adaboost, etc. Particularly Multi-View SVM which inculcated various dimensions or views of the information in the process of model training. It has previously been used for image processing tasks for “exploiting” different aspects. Similarly, it can be used for Text classification. This approach is not straightforward like a key-based approach and considers different aspects. Moreover, it doesn’t create a bias like source metadata detection.

Avenues where we need the hate speech detector

• Obviously, on top of the list stands social media platforms. Already our eminent players in the field are incorporating it into their interfaces. Facebook and Twitter have added this feature which allows the user to turn on this filter. Facebook is trying to apply a similar algorithm to various languages apart from English. Smaller platforms are yet to integrate it, it’s more like “they-have-to” when they want to expand to varied communities and build a healthier environment for the masses.
• The current rife has led us to depend heavily on E-commerce platforms. Often profane content is posted there as well, especially in the comment sections where anybody can write literally anything about anything and everything.
• Then we have collaborative information products like Wikipedia, StackOverflow where interaction between the users is not restricted.
• Online streaming platforms like youtube, Netflix, Hulu, etc also need to regulate their content.
• Messaging platforms like WhatsApp, messenger have been providing a room that invigorates the vigilante groups, leading to the spread of hate speech and violence.
• E-learning platforms which have become crucial for remote learning.
• Not to forget the gaming platforms, where the majority of youth spend their time.

Profanity filters are the most easily available and accessible technology for companies and businesses that aim at nurturing an efficient and amicable leeway in their network.

The mechanism on which Profanity Tool works-

First, it starts with scanning UGC (User Generated content) and filters the offensive “essence” in the online interactive community spaces. It is built on machine learning models which detect profane phrases or words. After the detection of these phrases or words, irrespective of the context they are then replaced with something like this @$&!.

Precis

With the rapid immersion of ourselves into the virtual spaces, we need profanity filtering to ensure these virtual spaces are secure and harmonious. For businesses, which have a fast-growing user network it is essential to integrate tools like profanity filters into their platform. This would make the overall experience of the user more amicable.

About T.F.L.W — The Four Letter Word

As mentioned earlier, undoubtedly there is an increase in our dependency on virtual networks so it has become indispensable to use a profanity filter for all the platforms.

T.F.L.W is formulated on a Machine Learning model that detects the inappropriate content and flags it, which further can be removed or replaced accordingly. Apart from Text detection this filter is compatible with voice recognition, it can detect audio that is abusive too. The results are fast and accurate, models are enhanced and precise detection is done within no time. T.F.L.W provides an audio detector as well. With remote working and unavoidable online meetings, it’s safe to have an audio filter along with a text filter. The user can use a mic for the tool to capture the audio and process it. It senses the abusive words and that’s what makes it unique compared to other profanity filters available.

T.F.L.W has achieved a success rate of 90% and ensures that the user has a pleasant experience. Companies and businesses with a burgeoning user community can integrate this tool into their platform. By incorporating this tool into your platform, you can enable all the features. Subsequently, the users can engage with all sorts of online activities without having to agonize about the unappreciated or offensive content.

The T.F.L.W tool is currently free and accessible to all. This framework eventually counters the toxic environment thus transforming our virtual spaces into came lots.

Reference:

Stephens, Richard, et al. “Effect of Swearing on Strength and Power Performance.” Psychology of Sport and Exercise, vol. 35, 30 Apr. 2020, pp. 111–117., https://doi.org/10.1016/j.psychsport.2017.11.014.

MacAvaney, Sean, et al. “Hate Speech Detection: Challenges and Solutions.” PLOS ONE, Public Library of Science, 20 Aug. 2019, https://journals.plos.org/plosone/articleid=10.1371%2Fjournal.pone.0221152.