#BotConAfrica2016: Praekelt launches new multilingual chatbot
Feersum Engine is a chatbot that can engage people in conversation using natural language processing and machine learning. Tipped for exponential growth, chatbots and conversational interfaces are already being adopted by many banks and financial services around the world, and their applications are becoming more varied as designers better understand the technology.
Streamlining interaction
Locally-built Feersum Engine brings artificial intelligence and conversational interaction to large audiences for customer support, brand marketing or public service outreach.
One use case for Feersum Engine, for example, is to cut down the number of simple to answer requests that overwhelm support and service agents, and is claimed to be its flexible enough to work in any environment from product support to government healthcare initiatives. By streamlining interactions and answering basic questions, the chatbot frees up human agents to focus on problems that need their expertise.
“In a call centre environment,” says Praekelt’s product director Belinda Lewis, “Feersum Engine can reduce the volume of calls agents have to deal with and increase the amount of time they are able to spend with customers who have real problems.”
It can chat over any messaging medium including email, Slack, Facebook and Twitter, WhatsApp and SMS. By drawing on Praekelt's JuneBug platform, users can chat over whichever platforms clients choose.
“We believe that Feersum Engine is a perfect solution for frontier markets like South Africa,” says Lewis, “where customers don't want to have to use their limited data allowance to download new apps that they aren't already using to interact with brands.”
Language and voice recognition
The chatbot is also designed to be multilingual, and can interpret questions in local African languages as well as international ones. By leveraging open source technologies to interpret natural language questions, it will assess and guide users to the information they need fast. Over time, its machine learning artificial intelligence will only speed up this process, spotting common problems and patterns in user queries and conversations.
Sentiment in replies can also be detected. This way it can pass the conversation off to a human agent if a customer is upset, for example.
“Feersum Engine isn't just designed with call centres in mind,” says Lewis, “It can conduct conversations over email, for example, to carry out stock checks or orders in an automated fashion.”
Ultimately, Feersum Engine will include voice recognition in its conversational toolkit, which Lewis believes will make it as viable in areas of low literacy as it is for desktop interactions.
