City guide Time Out has developed a conversational app for Google Assistant, used in Android phones, iOS and Google Home devices, through which people can ask up to 100 questions about what to do in their city. People can ask questions like, “What’s a good bar to go to tomorrow?” “What’s happening in London tonight?” or “Where’s a good place to go for cocktails with the girls this weekend?” Time Out’s assistant responds with three suggestions chosen by the publisher’s editors.

In time, Time Out will incorporate e-commerce deals into the assistant. For instance, people could hear three different theater show options, one of which could have a discount available as part of a partnership with the publisher. Time Out’s app for Google Assistant went live last week and is only available in London for now.

“This is highly applicable to what we already do with local businesses,” said Christine Petersen, CEO of Time Out Digital. “We already have a pretty broad offering with local shops and restaurants. Audio will play into this; it will be a really important part of the ad toolkit that we offer.”

Aware of the limits of a business reliant solely on advertising, Time Out began focusing on growing e-commerce and affiliate relationships in 2016. In the first half of 2017, revenue from e-commerce was up by 51 percent from the same period in 2016, according to the company. ESI Media’s London news brand, London Evening Standard, is another publisher looking to incorporate offers from local business like theater ticket sales into its Amazon Echo Show skill.

As voice-activated platforms grow, publishers and brands are spending more time thinking about what their brand sounds like. The female British voice of Time Out’s assistant is one of Google’s rather than one the publisher recorded. After each suggestion, the assistant asks, “Are you keen?” a slightly more British way of asking questions, according to the company. The assistant understands and responds to a few other playful questions. For instance, if the user asks, “Where are you from?” the assistant responds with the opening rap from “The Fresh Prince of Bel-Air.”

The cheeky tone of Time Out’s assistant is thanks to input from dozens of editors.

In early July, Time Out released an Amazon Alexa skill that delivers the top three things to do in a city, chosen by editors, but Time Out wanted to offer a skill that goes beyond the simple call-and-response model commonly found on the Echo and Google Home devices.

“We wanted this to be a human-focused project. We spent a lot of time in a room watching people ask questions to their phone,” said Sanjay Vakil, Time Out’s director of product development. “The rallying cry from our team was for it to sound like your best friend in the city who knows what’s going on, rather than a dumb robot.”

Vakil said Time Out’s app for Google Assistant took roughly three months of development work from a primary team of three people spending half their workday on the project, as well as input from dozens of editorial staffers to perfect the conversational tone. It was built on Google’s Dialogflow, which uses natural language processing to understand voice input.

While developing the assistant, the team worried about people asking it certain types of questions, including questions it would have no way of knowing, like the cost of tea in China. In this situation, the assistant provides a generic response. Another concern was when it doesn’t know the answer to what the user believes is a reasonable question, like, “What train will get me to Winter Wonderland in Hyde Park?” In this instance, the assistant offers help, like sending directions or a map to the user’s phone. When the assistant doesn’t trust its answer because something has broken technically, it offers a humorous reply, like, “Sorry, I don’t have any ideas for that. I’m still new at this whole assistant thing. What else do you want to do?”

According to Vakil, the data Time Out gets from Google is limited to reams of text and anonymized information about the questions asked. But in time, Time Out will be able to infer information on what people are interested in knowing and when. For instance, the assistant could offer the top three local kebab shops late on a Saturday night.

“Voice is a channel that will grow, and we want to grow with it,” said Vakil. “We felt we were on the cutting edge because we weren’t asking questions like, ‘When is my flight due?’”

  • LinkedIn Icon