Google LLC today announced some updates to the artificial intelligence technology that powers its Contact Center AI service.

Based on Google’s Dialogflow and Cloud Speech-to-Text tools, that service enables companies to create virtual agents that can handle many basic customer requests and assist their human customer service teams.

Google’s Dialogflow is a cloud-based service based on technology that the search giant obtained through the acquisition of API.AI Inc. in 2016. Made generally available in April 2018, the service allows enterprises to create virtual assistants without having to build from scratch the complex machine learning technology such projects normally require.

Google said Dialogflow is being boosted with a new feature called Auto Speech Adaptation in beta that’s designed to improve the contextual awareness of virtual agents. The idea is to help chatbots understand speech more accurately and respond correctly by taking into account the context of the conversation in the same way that humans do.

“Knowing context can help virtual agents respond more accurately,” Google product managers Dan Aharon and Shantanu Misra wrote in a blog post. “[For example] if the Dialogflow agent knew the context was ‘ordering a burger’ and that ‘cheese’ is a common burger ingredient, it would probably understand that the user meant ‘cheese’ and not ‘these.’”

Cloud Speech-to-Text, meanwhile, is Google’s transcription service, which relies on machine learning to accurately transcribe human speech into text form, therefore providing an important capability for most virtual agents.

For developers, Google offers a useful tool called “SpeechContext parameters” that can be used to add contextual information to improve the accuracy of their transcriptions. The idea is that developers can fine-tune their virtual agents in order to boost recognition of common phrases, such as a company’s product names.

Google is now updating SpeechContext with the ability to add “classes,” or popular concepts, to virtual agents so as to improve their understanding of what can sometimes be confusing phrases.

For example, if a caller says something that Cloud Speech-to-Text transcribes as “It’s twelve fifty one,” SpeechContext classes can be used to refine the transcription in several different ways:

In addition, SpeechContext is getting further capabilities, including a new “boost” feature that helps to reduce the number of false positives, which occur when a phrase wasn’t mentioned but appears in the transcript. It also gets a new feature called “phrase hints,” which developers can add to increase the probability that commonly used words and phrases specific to their business are understood and captured by Cloud Speech-to-Text.

Constellation Research Inc. analyst Holger Mueller told SiliconANGLE that Google is simply trying to make it easier for developers to embed Dialogflow and Speech-to-Text into their chatbots and contact center applications.

“Bridging these channels is key for building these kinds of next-generation applications,” Mueller said. “Only a single text body and conversational flow is needed, creating key consistencies that speed up the creation of these apps.”

Images: Google