Google Delivers Slew of Contact Center AI Updates

almost 3 years ago by Lucy Cinder

Google Delivers Slew of Contact Center AI Updates

In a world where customer experience is crucial to business success, the contact centre exists at the heart of every crucial business conversation. In the age of digital transformation, access to the right technology can make or break the success of the average contact centre, helping organisations to implement the best state-of-the-art solutions for customer care.

In July 2018, Google’s Cloud team announced the arrival of Contact Center AI – a solution to help companies apply intelligent solutions to their contact centre offering. Now, a year later, the GCP is introducing a host of exciting new updates to the tech at the centre of Contact Center AI, particularly Cloud Speech-to-Text, and Dialogflow.

Enhancing the Value of Virtual Agents & Auto Speech Adaptation

One of the first updates introduced by Google’s Contact Center AI this month appears with the speech recognition features of virtual agents. As virtual assistants continue to support today’s employees in delivering stronger around-the-clock user experiences, they’re becoming increasingly important to the modern contact centre. However, automated speech recognition struggles in noisy contact centre environments. Fortunately, Google is introducing a new feature to help virtual agents understand what customers need.

Auto Speech Adaptation is the new intelligent solution from Google that brings context to the customer conversation. Speech Adaptation describes the learning process that Google’s virtual assistants use to get to the heart of the nature of a conversation. For instance, the Dialogflow agent might now that the context of a conversation is a customer ordering a burger. Because of that, the virtual bot would be able to understand that the customer probably meant “bun,” not “run” or “fun.”

Additionally, because an agent would understand that “mail” is a common word to use when discussing a return, the bot would be less likely to confuse the word with “male.” The new Auto Speech Adaptation feature ensures that virtual agents can take all available information into account when processing conversations, leading to a 40% or higher increase in accuracy. Auto Speech Adaptation will be turned off by default, but admins can switch it on within the Dialogflow console.

Cloud Speech-to-Text Improvements

Another significant update that Google has made this month comes in the form of new Speech-to-Text baseline model improvements for phone-based virtual agents and IVRs. In April of last year, Google introduced new pre-build models for transcription of the video and phone calls. In February this year, those models became generally available. Today, the models are more advanced than ever, with a 15% improved accuracy rate for US English. The introduction of speech adaptation can also help customers to achieve even greater accuracy levels.

Accurate transcriptions in the contact centre can make it easier for agents to respond to the requests and needs of today’s customers. The updates improve the quality of transcription accuracy for human agents.

Additionally, developers in Cloud Speech to Text will typically use “Speech Context” parameters to provide additional information to that transcription. This process improves the agent’s ability to recognise phrases common in a specific environment. Today, the manual speech adaptation tuning in the Speech to Text environment will be richer than ever, with new enhancements in the Dialogflow and Cloud Speech-to-Text APIs. Google has announced the arrival of a further “boost” feature that will allow developers to use the best possible speech adaptation strength to suit their use case. This should help to increase the likelihood of complex phrases being captured.

What’s more, as part of the upgrade for SpeechContext, Google has also announced an expanded phrase limit. This supports the phrase hints that developers use to improve the probability that commonly used phrases or words related to a business or virtual can be captured by the ASR. The maximum number of phrase hints has been increased by 10 times, to 5000, allowing for the optimization of thousands of jargon and industry-specific terms.

Google has even introduced the arrival of SpeechContext classes – the pre-built entities that reflect common concepts and provide Speech-to-Text with more context for accurately recognising and transcribing speech input. Using classes will allow developers to tune ASR for a whole list of words at once, rather than just adding them one by one. Various classes are available to provide context around the sequences of numbers, addresses, digit sequences, and money types that appear in the conversation.

Improving the Cloud Speech-to-Text Experience

Since the Cloud Speech-to-Text offering was introduced around 3 years, long-running streaming of the service has been a common demand among Google Cloud customers. Up until very recently, the Speech-to-Text offering could only support streaming audio in increments of one minute at a time. However, this was problematic for long-running transcription use cases like live video, meetings, and phone calls. Going forward, the session time is increased to five minutes. What’s more, the API will allow developers to continue from where the previous streaming session left off with each session, meaning that live automatic transcription is virtually limitless.

Google is also introducing the arrival of Cloud Speech-to-Text support for mp3 files. Up until recently, Google had supported 7 different file formats for Speech-to-Text transcription, but processing MP3s required expanding them into a different form, which meant extra work for business users. There’s now native support available for MP3, with no conversions necessary.

Already, Google has revealed positive case studies from customers who are taking advantage of the exciting opportunities available from the Cloud Contact Center AI. One of the largest retailers in Australia, Woolworth’s recently announced that it had achieved significant business improvements using the Auto Speech Adaptation feature. According to Nick Eshkenazi, the Chief Digital Technology Officer, Auto Speech Adoption allows Woolworths to respond more accurately to customer queries. Previously, the company used to need months to create high-quality, IVR experiences.

Google is thrilled to be announcing these new updates to the Cloud Contact Center AI solution to provide users with better speech recognition and transcription. The company plans on continuing to roll out new upgrades and developments to improve customer experiences for contact centres of all shapes and sizes around the world today.

source uctoday

Industry: Unified Communications & Artificial Intelligence