The Rise of Hate Speech in Social Media: How to Address It in Any Language
It’s a fact that bullies and hate speech mongers have been around since the beginning of civilization. With the rise in technology, hate speeches have become widespread. This is because the internet offers a layer of anonymity that was absent before.
If statistics are true, 4 out of 10 Americans have experienced some form of online harassment and 62% consider online harassment a serious problem. A 2017 survey states that over one-quarter of online harassment victims in the US said they stopped, reduced, or changed their online activity because of online harassment. While 41% of internet users had experienced varying degrees of online harassment, 28% reported they had gone through harsh forms of harassment. This included physical threats, stalking, sexual harassment, and sustained harassment.
While the above statistics are for the US and represent the English-speaking population, social media users across the globe aren’t all posting in English. This means the problem of non-inclusive language and hate speech needs addressing in various languages.
Each day there are 500 million tweets shared, 350 million images uploaded to Facebook, 720,000 hours of content uploaded to YouTube, and 100 million photos and videos uploaded to Instagram. That sheer volume of content online today, along with the expectation that offensive content must be dealt with rapidly requires addressing by any brand that publishes and manages digital content.
WATCH NOW: Using AI To Protect Your Global Brand: Tackling Non-Inclusive and Hate Speech. This on-demand webinar features Welocalize VP of AI Innovation, Olga Beregovaya and Emil Atanassov, VP of Internationalization at ServiceNow.
How Widespread Is the Problem of Non-Inclusive and Hate Speech in Social Media?
Borrowing an expression from the technology industry, non-inclusive language, hate speech, and harassment are now a feature of life on the internet. Non-inclusive language and hate speech in their milder forms create a layer of negativity that all internet users must comb through every day. In its most severe form, it can compromise the user’s privacy. It forces them to choose the platforms they would like to participate in or which ones to avoid. It can even pose threats to their physical safety.
The above data makes it clear that non-inclusive language, harassment, and hate speech are becoming synonymous with social media.
Negative Impact of Online Non-Inclusive and Hate Speech Content
The harmful impact of non-inclusive and hate speech content online ranges from mild negative feelings to long-lasting physical harm. Name-calling and harsh comments are the most common sources of online non-inclusive language. However, this isn’t the end. Riots, mob lynching, and other hate-related crimes have been attributed to the misuse of social media platforms.
Setting Standards and KPIs to Measure the Quality of Offensive Content
To differentiate between non-inclusive language and hate speech and to measure the quality and severity of the content, you need to set standards and key performance indicators (KPIs). The set standards should account for language barriers. They should also be sensitive to the culture, region, and community.
People post an immense amount of data online every minute. Processing every tweet or every post through a human moderator isn’t possible. This is where artificial intelligence (AI), machine learning (ML), and natural language processing (NLP) can take over. AI algorithms flag the content based on keywords and language. KPIs gauge the severity of the content so it can be categorized under non-inclusive language or full-blown hate speech.
The greatest advantage of using the combination of Al, ML, and NLP is you can set up an engine that learns while it is processing data. To address the menace of non-inclusive language and hate speech in different languages, AI needs to be programmed to understand the language. For example, you can program AI to understand French or German by giving it keywords in French or German. AI is processing data in locale vernacular languages and not just English. For an AI-based system, its learning is a continuous process.
Reliable Solutions to Identify and Monitor Offensive Social Media Content
Non-inclusive language and hate speech can tarnish any social media platform, brand, or individual. It also has the potential to cause a long-lasting negative impact on individuals and society. The most reliable solution to scale down negative content on social media platforms is to use technology. AI, ML, and NLP, along with human intervention, when needed, can help social media platforms identify, flag, and monitor offensive content online in any language.
By leveraging AI as part of an overall language solution as well as personnel around the world, Welocalize can monitor and rate multilingual content, and implement processes to remove offensive content.
The solution analyzes client products, marketing materials, user assistance materials and identify offensive/hate speech in non-branded, user-generated content (UGC) such as knowledge bases, forums, opinion portals, and emails.
In addition to understanding UGC that may cause harm, the Welocalize brand also considers the impact that offensive content may have on people working with it. We have multiple programs in place to support the health and well-being of our team.
We have 60+ pre-trained language models to identify the sentiment of your content and a growing team of linguists to help process content. These models also suggest alternatives to make the language more inclusive and flag hate speech. Welocalize can help your business be sensitive and inclusive. Contact us today to learn more.