Up until now, Debbie, our enthusiastic AI receptionist, has only been active on debbie.chat, WhatsApp and Facebook Messenger – all text-based mediums, created in Google’s virtual agent designer Dialogflow CX. To get Debbie on Alexa? Turns out translating text to voice is not as simple as A, B, C…
As well as all the intricacies of language, explained more in our previous post, we were faced with the recurrent demand of cross-team communication; how designers and engineers articulate their ideas is reminiscent of chalk and a popular milk-based product.
Added into the mix was a long-standing dream of Hi9, to create our own ‘vertical’ – a platform which produces, supports, markets and sells their product or service, as opposed to a ‘horizontal’, which aggregates across the business in one particular aspect. For example, Direct Line VS CompareTheMarket.
So, how did the creation of DeeDee manage to achieve all three of these aims?
Before DeeDee, building a ‘skill’ – or app – for Alexa relied on creating a set of ‘intents’ – what you think the user might say – and this happened in the Alexa Developer Console, whilst the rest of the dialogue was embedded in the code. Without being proficient in the JavaScript or Python coding languages, after writing the script, designers were reliant on engineers to build the bot for them.
Hi9 is always keen to combat ‘technical debt.’ This is what you’re left with when you prioritise quick fixes over long-term solutions. It would have been relatively quick for our engineering team to code Debbie into Alexa, but in the long-run it’s more efficient for the design team to be able to connect the script to the code directly. Beyond learning to code, how could we achieve this?
We needed a functional middle-ground between the designers and engineers. By communicating regularly to ensure alignment and understanding between the teams, we already had conceptual middle-grounds. Now we needed somewhere that both teams could collaborate on the same project, that utilised more of the design team’s skills in order to contribute to the engineers’ tasks, all whilst accommodating each team’s preferred ways-of-working.
And so, DeeDee was born! A visual tool for chatbot design which uses different types of blocks to build a conversation. The design team are able to insert sections of the basic script into the relevant dialogue blocks, add expression blocks to anticipate what the user will say, and connect the blocks to ensure the smooth flow of conversation. The engineering team are able to access these same blocks, add Speech Synthesis Markup Language (SSML) to transform Alexa’s voice into Debbie, and use a ‘webhook’ block to connect the conversation to the necessary external transactions, such as extracting data, making calculations or sending an email.
Not only have we created a functioning tool enabling language translation between text and voice, where the design and engineering teams can work more efficiently together, but DeeDee is self-reliant. Whether it’s Google or Alexa, or in the metaverse, DeeDee will soon be flexible enough to accommodate whatever platform we need to be on. Easy as 1, 2, 3.