Not Just Voice: The Natural Language Revolution

According to widely circulated estimates, and recently rounded up by DialogTech here, voice technology is set to explode this year. Half of all online searches will be voice searches. Thirty percent of web browsing will be screenless. Voice interactions through smart home devices and mobile voice assistants will guide how a large number of consumers navigate through their digital lives.

The growth of voice search is a behavior change, as more users adopt voice interfaces. Google finds that 52 percent of households that own voice-activated speakers keep them in the living room, indicating that they are being used in a communal space for general purposes. (The rest of the households are split almost evenly between keeping these devices in the bedroom or in the kitchen.) Even if only a quarter of households in the US actually own these speakers currently, this shift in behavior has wide implications for marketers.

The transformation that marketers must oversee in strategy isn’t necessarily rooted in voice, however. Successful brands engage with their customers in an omnichannel way, so marketers must accommodate voice and how it changes the other channels. Being voice-first doesn’t mean that all brands are giving up screens completely. A quick visit to almost any QSR chain will confirm that in-store customer experience is getting increasingly visual and screen-based with touch-screen ordering.

What’s really at the heart of the movement changing customer interactions is Natural Language Processing (NLP). Considered a kind of AI, it includes, yes, the speech recognition that enables a voice assistant to respond to a talking human, but only if it understands the use of the words that it hears. Being able to interact in a natural way is what NLP looks to achieve, making it a valuable marketing tool that can attract and influence customers.

Marc Ferrentino, chief strategy officer for search experience platform Yext (they work with big chains like Denny’s and Taco Bell) told me that NLP will become “the dominant way humans interface with computers over the next few years.”

A key moment a last fall came with a new Google paper on BERT, which stands for “Bidirectional Encoder Representations from Transformers.” This neural network-based technique, introduced and open-sourced by Google in 2018, focuses on understanding all queries, including those it hasn’t seen before. Of the billions of searches submitted each day, 15 percent are new ones.

The paper, written by Google fellow and vice president, search, Pandu Nayak, explains how this technique draws on the presence of all words used in the query to provide context and to understand the intent behind the query.

One trend last year that indicates BERT’s success is that queries are getting longer. Users feel more comfortable using longer, more natural questions because they now yield productive results. This use of natural language replaces what the paper calls “keyword-ese,” the tendency for searchers to just type in (or say) the relevant keywords, without trying to formulate the question they hope will be addressed anyway, somewhere down the search results.

“What we’ve seen,” Ferrentino said, “is an increase in multi-word queries. People are leveraging more and more questions, using search engines to search in questions.”

He added, “Seventy percent of searches are three keywords or more. They’re not always full-form questions, but that’s mainly one of the big reasons why searches return results now.”

If users are more likely to speak to a voice assistant in complete sentences, this will also change how they type out search queries. Either way, NLP techniques will gain fluency. And this raises the standard across other CX interactions.

Ferrentino expects that with increased natural language proficiency, “the entire martech stack will transform because of it.” Marketers will have to reevaluate which solutions they use according to how well they handle natural language.

“The first area natural language has disrupted is search,” he stated. “Expectation of this experience is bleeding beyond search, and we expect that to continue over the next few years. This year will be the year for marketers to begin to think about a natural-language structure.”

NLP will also have an application for conversational interfaces, but its impact on chat is “future looking,” according to Ferrentino.

He sees search developing in a conversational way, as well, following a “Q&A” format, where the AI and the user ask follow-up questions to further refine the Google result. And this kind of interaction will also be valuable between brands and customers.

“Over time, when you start to see what the interactions look like, it creates a virtuous circle, and you will know your customers better,” he said.

Ferrentino predicts we will see the rise of the knowledge graph that structures data using the natural language of its customers, “building up the brain of the brand.”

Brands will know more about customers, and not just about the intent behind the questions they ask, but also their intent to purchase.

“The more people are searching, the more that intent marketing will be part of a CMO’s playbook,” he said.

Direct-to-consumer (DTC) brands have gained ground through alternative, digital channels and by building meaningful relationships with their customers in these interactions. Digital marketing and growth platform AdRoll works with many DTCs.

AdRoll President Scott Gifis has been watching the development of conversational tools, and sees the opportunity for brands to gain a deeper understanding of their customers. But, marketers need to be patient for the technology to develop, and for customer relationships to grow over time.

“Conversational marketing powered by AI has seen some interesting revolutions in recent quarters,” Gifis said, adding that because of dynamic websites that provide personalized experiences at scale for individual customers, the field has seen “a resurgence in personalization.”

He explained, “Marketers jumped the shark with personalization engines because we lost sight that we’re focusing on a relationship, and instead focused on the near-term relationship of a conversion.”

He stated, “Conversational chat tools (bots) are prevalent, but no clear ‘winner’ has emerged as of yet. That said, we are seeing the idea of conversational commerce more and more widely adopted, with hybrid human support supplemented by conversational tools, and expect to see investment in here increase in 2020. DTCs that deploy these tools not only see higher conversion rates and cart values but also glean deeper insights into their customers through these interactions that pay dividends beyond short term ROI.”

WitLingo, a CMS for voice-first experiences, connects brands with Amazon Alexa, Google Assistant, Cortana and other voice searches.

The company’s founder and CEO, Ahmed Bouzid, worked on Amazon Alexa, and builds his technology to navigate the “real world” without interruption.

“With the emergence of the smart speakers, voice is delivering on its true value,” Bouzid said. “[Users are] doing things without having to touch or view something. It’s touch free. You could be folding laundry, talking to somebody and be able to engage with the intelligence – ask a question, like ‘what time is it?’ while continuing the conversation.”

When moving around in the world, you can see an ad and ask a question about it. WitLingo’s clients also use product packaging as a call to action to bring customers into the voice-first portal.

“Voice is ephemeral,” Bouzid stated. “You may want to offer a customer to get a coupon, look for more information, or have [the brand] text the information to you. It becomes a bridge to information you can follow up on. There’s minimal friction in that.”

The prompt could also refer you to a call center for customer care.

“The core best practice is to be concise, to the point, to be useful and to have a sense of context. Brevity is the key thing.”

The second key practice is that the message is actionable.

“What makes voice interesting in this format is that it connects you in the real world without having to go to the digital world,” he said. “You stay holding that piece of paper. That’s the thing that is compelling about it.”

In the years to come, marketers will have an even wider range of engagements to choose from, rooted in voice and NLP. They will be able to keep things concise while their customer is out in the world, and when appropriate, build deeper connections through natural conversation.

Total
0
Shares
Related Posts