The machines are getting better at understanding you

It’s never fun to feel misunderstood, but when you’re trying to interact with a bot in place of a human and getting nowhere, that’s a whole other level of frustration.

Over the last decade, companies have increasingly turned to chatbots to make customer service more efficient. For the most part, these chatbots have used a technology called natural language processing, or NLP, which uses machine learning to better interpret what humans say. 

But the uneven experiences in customer service can be attributed in part to the shortcomings of NLP. “Historically, the idea was that a human language technology stack looked like a pipeline,” says Dzmitry Bahdanau, a research scientist at ServiceNow. “First, you note the part of speech, then the syntactic structure, then the semantic structure of the sentence.” This leads to a basic sense of the customer’s intent or state of mind, he says, which helps to classify queries and analyze sentiment. But it’s hardly enough to carry on a complex conversation.

This limited ability to understand human language gave rise to a desire for a more natural way for computers to interact with humans—which led to the creation of much more powerful chatbots using generative AI, for example.

These new tools use natural language understanding, or NLU, which is trained to bring a much more nuanced sense of context and sentiment, and is more capable of having natural conversations with end users.

“NLU helps provide a much better end-user experience,” says Sibo Ding, principal product manager at ServiceNow. “The chat experience is more natural, and it has a better accuracy.” 

Because of this, virtual machines are better able to understand us, and that in turn is creating new opportunities for companies to have better automated interactions with customers using AI.