How to Make Your Chatbot Sound Natural
We have all grown accustomed to technology — it’s all around us and we can’t even imagine our future without it. As we are moving forward, interaction devices evolve and interaction design progress. Thanks to science fiction books and movies we know exactly how we will interact with the machines — we will use voice commands, sweeping gestures, and virtual screens, just like Tom Cruise once did:
Interaction, where you use such forms of communication like touch, voice, and gestures, feels easy to us. It doesn’t require learning, and in computing, this UI is called natural user interface, or NUI. The word natural is used because interaction equivalent to our human behavior. From our first days of life, we’re grabbing and moving things, trying to talk, like the people around us. And it is perfectly natural that these interaction methods are mirroring in the human-machine interaction.
Bill Buxton, a principal researcher at Microsoft, once said:
NUIs exploit skills that we have acquired through a lifetime of living in the world, which minimizes the cognitive load and therefore minimizes the distraction.
He also claimed that voice user interfaces (VUI) are probably the most natural user interface, especially when driving a car — obviously, your arms are tied to the steering wheel and your eyes are supposed to be on the road. Voice is cool because you can easily transfer a great amount of information. Evolution made it the most effective tool for communication, and modern speech technologies made it possible for the machines to talk and what’s most important, to listen.
Top 3 Most Popular Bot Design Articles:
Voice is a common human skill so you can assume that your users already have that skill because they are human. Next challenge here is for VUI developers who create dialogs/skills/actions, teach digital assistants to communicate and make them useful. The challenge is pretty rough because in order to understand user’s intents a machine has to be aware of the context. In order to sound natural, there should be personality, and etc. We’ve made a short list of handy tips that you have to keep in mind when creating chatbots and AI-powered assistants.
Context
As humans we use context in a natural way, we barely think of it. We mechanically know how to arrange conversation for different people or places. We speak to our kids, parents, friends, and colleagues in various ways. We speak loud and bluntly at home, and we keep the conversation levels down when we are in public. Chatbots and virtual assistants have no such indigenous knowledge. That’s why this information should be given. A prior context might include information about time, location, and user’s device. A posterior context includes previous queries/answers, name, authorization data, the mood of the user and assistant itself. Users should be greeted considering the time of day. Just tell them their name, don’t ask things you already know, don’t guide an experienced user as a newbie and get ready for ellipsis.
Personality
To sound natural digital assistants have their own distinct personalities. For instance, Alexa is playful and she has her own opinions about all kinds of things. Moreover, her opinions and preferences vary from country to country: her favorite beer in the U.S. is different than her favorite beer in Germany. When Amazon developers started creating Alexa, they simply wanted her not to sound like a heartless machine. But they were stoked about how many people fell in love with her personality. You may read about this incredible experience in more detail here. Now users are expecting Alexa to reply and behave accordingly to her personality, and her character integrity contributes to the increase of their trust. This principle limits the ability to generate responses automatically, but it is crucial to user experience.
Conciseness
Short wording reduces cognitive load, saves time and sound natural. Shorten your texts to the benefit of really important info and leave out facts and instructions your user already knows. In case your assistant has a display, you can keep some pieces of information off the radar.
Flexibility
Use NLU (Natural Language Understanding) resources to understand all the possible query formulations and probable errors of the speech recognition. You must foresee the user’s possibility to change the dialog at any point – for example, in case they want to refuse the order or to provide the information in a form suitable for them. Also, remember that the user may give more information than you ask for. For example, when asking ‘Would you like to order Margherita pizza for $13?’ get ready for the answer ‘Yes, two of these please’.
Naturalness
Assistants’ speech has to be natural. Avoid repetitions and bureaucratic language. Where possible, use implicit confirmation and active listening techniques — repeat back key parts of what the user has told you. And don’t forget to duly handle greetings, goodbyes, thanks, help requests and etc.
Proactivity
Avoid situations when a user doesn’t know what to do next. Monitor blocks in dialogs and think of a way out. Carry on a dialog by means of questions or coach marks, managing the user’s expectations. Suggest buttons here is a fine way.
In practical terms conversational design is confined with technologies’ development level, implementation features for each specific assistant, and service interfaces. But following these tips will help you to create an easy conversation. This field is relatively new, we all learn from own mistakes, so don’t be afraid to make ones. And remember — a good dialog is a natural dialog.
Please, feel free to contact us if you need help with the conversation design or chatbot/smart assistant development — we at Just AI are always happy to help 😌
Let’s build amazing humanlike experiences together!
Don’t forget to give us your 👏 !
