Over three years, SoftBank, a Japanese business, has created an artificial intelligence filter that can recognize furious voices and automatically translate them into calm speech.
A SoftBank developer, Toshiyuki Nakatani, said, “We developed the emotion suppression system to address the social problem of customer harassment of call center employees and to protect them.”
Two steps make up the voice filter’s working. Initially, the AI recognizes an angry voice and summarizes the speech’s main points. It uses acoustic techniques in the second step to change the voice into a more pleasant and natural tone. Surprisingly, the filter softens the tone and does not alter spoken words. Call center employees will, therefore, continue to receive insults, but they will do it in a kinder, less upsetting way.
SoftBank’s developers recorded over 10,000 voice samples from 10 actors to train the AI. These recordings included at least a hundred common expressions, including screams, accusations, threats, and demands for an apology. Thanks to this large dataset, the AI was able to learn how to efficiently convert aggressive speech into calm speech.
The potential benefit to the emotional well-being of operators is encouraging, even though it’s yet unknown when SoftBank will use this scream-filtering AI in its call centers. SoftBank’s AI filter has the potential to significantly lessen workers’ stress and anxiety by converting aggressive conversations into more peaceful ones.