ToxMod is a powerful, intelligent voice moderation tool that is already used by studios around the world to combat toxic behavior in voice chat. Now, we’re expanding ToxMod’s capabilities beyond English to support 17 additional languages, enabling developers to protect their players no matter where they are or what language they choose to speak.
The road to understanding toxicity in multiple languages
Creating the best, most informed toxicity prevention in a new language isn’t as simple as integrating some form of machine powered language translation like Google Translate. We often say that ToxMod goes far beyond mere text transcription, which can only handle rudimentary keyword-matching and lacks the ability to understand the context of a conversation. Fighting toxic behavior is more nuanced, and requires fluency in a language and the culture of the language’s country of origin — as well as the subculture and psychology of gaming and online behavior in general.
When developing for a new language, Modulate employs AI, language models, and – importantly – human experts who are fluent in each language and skilled at identifying toxic gaming behavior. Working alongside Modulate’s experts in AI and language models, they help us develop the best possible starting points from which to develop effective, accurate models for ToxMod as they expand into new languages.
All eighteen languages (including English) are available to customers upon request.
“We have a recipe we know works. And we have a responsibility to be accurate.”
In beta: Spanish voice chat moderation
Spanish (the second most commonly spoken language in the U.S.) is the first language that is available in beta to ToxMod customers who opt in. Right now, we’re labeling new languages in beta when we determine they have strong, internally validated coverage of all our harm categories, and have a high degree of confidence that the performance in that language is nearly up to the standard of our full release languages.
During this beta period, Modulate will continue analyzing Spanish voice data and, with the help of moderators and community teams, fine-tune ToxMod to better understand the nuances and intricacies of the way players speak in Spanish in live game voice chat. The feedback that studios, moderators, and community teams provide for our Spanish beta will be used to tune and update the Spanish models, as well as improve the ability for ToxMod to identify which language is being spoken and separate reports, audio clips, and categorization between different languages.
In alpha: Toxicity prevention in Portuguese and beyond
In addition to Spanish, we’re in the alpha phase for 16 additional non-English languages. Whereas our full release languages – currently English – have battle-tested coverage of all harm categories, and our beta languages have strong, internally validated coverage of all our harm categories, alpha focuses on catching the most egregious harms and rapidly expands to handle more nuanced scenarios. These will go through a period of rigorous training and testing prior to beta release, where initial data for the language is collected and analyzed to help lay the foundation.
Studios that have implemented ToxMod will have the opportunity to enable languages that are in the alpha testing phase. The only change customers opting in to a beta or alpha period for a language are some new options in their user interface. We expect Portuguese to move from alpha to beta phase within the next couple months.
Our full list of supported languages now includes:
- Chinese (Mandarin and Cantonese)
What multi-lingual support means for players
Until now, players in games that employ ToxMod’s proactive voice moderation have been protected from toxicity in English but not other languages, allowing toxicity and disruptive behavior to proliferate if not spoken in English. With our new multilingual feature set, ToxMod can now understand the full context of a conversation – even if it is a mix of languages! – and better escalate real harms that are happening in any language we support.
As studios roll out these languages in ToxMod, players who speak any of our 18 supported languages can now count on the same level of protection from harm and harassment – meaning better gaming experiences and happier communities. And players can feel assured that we’re following the same formula in perfecting all of our language models, including tapping fluent speakers in every language and working with moderation teams who are able to inform the Modulate team and our models about what harassment or harm looks like in each language.
What multi-lingual support means for studios
Enabling multi-lingual support is seamless and fits into the existing ToxMod workflow that moderators expect. ToxMod intelligently determines the language that each player is speaking, meaning that instances of voice chat toxicity are directed to the appropriate team member that is equipped to support each language.
ToxMod customers can rest easy knowing that they can now provide a much larger percentage of their player base with the level of community protection they’ve come to expect in English. Studios of all sizes, from indies to the largest AAAs, can protect their players across all the languages that ToxMod supports.
Multi-lingual support is the latest feature to bolster ToxMod’s enterprise feature set, alongside full 24/7 customer support, single-tenant ToxMod instances for multiple titles, and partnership with AWS and Nintendo.