Co-Founder Maria Krausch talks about voice assistants and data protection

How did you get the idea to develop language assistants based on artificial intelligence? 

In the beginning there was the idea to help elderly people who have difficulties using apps and programs. Semantic software enables easier access to technology through speech, gestures and writing. My co-founder Pàscal Casadei van Raamsdonk had already developed language assistants 10 years ago in collaboration with the University Tor Vergata in Rome. The many applications in the field of "conversational commerce" motivated us to create Wikifriend.  

Which technological methods does Wikifriend use to analyze the spoken word in order to react to questions and wishes in real time?

Speech recognition and speech processing have different nuances that are worth deciphering: Is the user in a good or bad mood? Which reaction of the speech assistant is inappropriate in which situation? Wikifriend has its own semantic algorithm that determines word types, carries out mood analyses, recognises certain connections and evaluates content. People use all these skills intuitively when communicating with each other. Wikifriend has developed a language analysis (NLP), which is especially aimed at the analysis and characterisation of the feelings of individual users. With the help of our Deep Learning algorithm, which is constantly learning, our users can have a personalised conversation. 

What makes your solution stand out from other language assistants on the market?

Companies want to improve their customer relationships and generate new leads by using a virtual assistant. But voice assistants can also frustrate customers if they give the same answers over and over again or remain completely silent. To be effective, a voice assistant must do more than just wait for someone to ask a question. The new generation of high-quality speech assistants has an innovative approach: With proactivity and well-structured scenarios, Wikifriend's software is able to proactively ask users questions, provide personalized answers, guide the dialogue, respond to their mood and even bring them back into the conversation.

How does Wikifriend protect the personal data of users? 

Other voice assistants like "Alexa" (Amazon) stream the conversation with their users into the cloud as soon as they say the "wake-up word". These "smart speakers" have come under fire because employees record and analyze excerpts of conversations. Wikifriend takes a different approach here: Customers want to use smart voice assistants, but do not want to be permanently overheard. This is why our software performs speech recognition and natural speech processing directly on the device itself. This means that no voice data leaves the device (on premise). In addition, the entire software is hosted in Germany.  

Source: https://blog.deeptechberlin.de/chat-emphatische-und-interaktive-sprachassistenten-von-wikifriend/