Case Study: Largest Online Game Developing Company achieves improved toxicity detection in gameplay audio with Pythian

Pythian Develops Prototype For Toxic Sentiment Detection For Online Game Developer

Largest Online Game Developing Company, one of the largest developers of multiplayer online games, needed a way to detect toxic behavior in gameplay audio conversations. Its existing toxicity models worked well for chat messages, but the company wanted to expand them to spoken in-game speech, so it turned to Pythian for help.

Pythian built a prototype using Google Cloud Speech-to-Text and TensorFlow to transcribe gameplay audio and identify toxic fragments, while also improving recognition of gaming slang and varied accents. The solution enabled the company to extend its toxicity detection from text to audio, reuse existing models on speech streams, and evaluate transcriptions against in-house labeled data to assess quality.