Chatbots: How to make a delightful conversational design
The rampancy of AI driven chatbots is imminent, for their promising ROI, which has been corroborated by successful deployment of conversational agents by companies like Mastercard, Spotify and Sephora. Chatbots, today, are created with the intent to automate customer service communication, provide an omnichannel experience, and augment sales & marketing.
However, while the grandeur of a successful conversational agent creates customer satisfaction, an unaccommodating chatbot with limited language understanding and a distasteful conversational design, can be just as ignominious for a business.
There are, in fact, several quandaries to consider while developing a business chatbot. Two key factors for your chatbot to be successful, is its Natural Language Understanding (NLU) and a robust conversational design. Today, we take a look inside chatbots, their types, and what truly allows them to have a robust and intelligent conversational design.
Outline
In this article, we shall have the look at the following :
1. How do chatbots work?
2. How good is your NLU?
3. Unhappy path handling
4. How to make a chatbot that excites users?
1. How do chatbots work?
There are two primary types of chatbots:
1. Rule-engine based
2. Powered by machine learning (subset of AI)
The rule-based chatbots are a lot simpler, and won’t understand complex commands and requests, or be able to discern context. However, machine learning driven chatbots are more advanced in terms of intelligence and adaptability. They can better understand context and react appropriately with the help of NLU engines and Dialog Managers.
Machine learning incorporates natural language processing to help chatbots achieve this, and as a result, can carry more sophisticated conversations than their rules-based counterparts.
Prior to development, Facebook Messenger and Slack are common places to build chatbots or we can host them locally on our website.
As soon as it’s live, the customers can begin interacting with your conversational agent or chatbot, which can answer questions, make purchases with actionable conversations, and more. The more your conversational agent interacts with people, the smarter it gets. That’s the beauty of Machine Learning!
2. How good is your NLU?
Natural Language Understanding (NLU) is a key component of a chatbot architecture, and uses Machine Learning to extract meaningful information from a user’s utterance.
Customers like a human experience, even when they’re dealing with a machine. In fact, customers tend to prefer chatbots that are so convincingly. That is why having a good language understanding mechanism is paramount.
Two aspects of a basic NLU engine are intents and entities. Intents are the different categories of utterances to classify what the user could mean while conversing with the bot. Whereas, Entities are important information present in an utterance.
For example, in the utterance — “I am looking for a Chinese Restaurant in the center of the town.”
Clearly, the user intends to search a restaurant, and has provided the cuisine of his preference. Therefore, the following can be concluded:
“intent”: “restaurant_search”
“entities”: {“cuisine” : “chinese”}
Key points to keep in mind, for anyone who is building conversational interfaces:
Go through all the intents defined in the scope of the Bot and be clear about each one of them. This ascertains that you don’t write common utterances for different intents.
1. Add new words in each utterance to widen the vocabulary, and help your bot learn variety is language.
2. Do some user research, and ask your friends about the way they would speak to the bot by giving them the particular intent.
3. Write distinct utterances in a particular intent but don’t lose the context.
It is important to have distinct utterances of each intent in your chatbot training data. Also, there must be at least 50 utterance examples to make sure we have enough test data to evaluate our model. Make your NLU ‘intelligent’ and ‘predictable’ — by making it contextual.
3. Unhappy path handling
Sometimes, users don’t ask relevant questions from the bot, or respond to them in the way they should. This leads to unhappy paths.
Typically, users will ask questions, make small-talk, change their mind, or otherwise stray from the happy path. Often users get generic responses to chatbot misunderstandings, “Oops! I didn’t get that” and variations of that, which are fallback messages to a request that chatbots don’t understand.
This type of generic responses exhibits a weakness in conversational technologies. If your bot fails, consider diverting the user’s focus away from the failure event, and restore some delight to the user’s experience. There are a lot of strategies chatbot development, and are already being practiced today to sustain engagement with their users.
- Get creative with Fallback messages.
If there are multiple fallback messages, rotate them so that any misunderstanding doesn’t seem like a message equivalent to an error 404. No user wants to hit a dead end, but if your fallback responses are creative enough, then you might just delight users with the unexpected. - Gracefully focus users
Sometimes, the communication breakdown with users, occurs in the middle of a conversation, and the bot is able to maintain context-states, then it can serve up two messages in succession. The first is the fallback message to the misunderstanding, and the second is a reminder of the context. In this sense, bot will be able to move the focus back to their goals and not let the misunderstanding become a distraction in your user experience. - Tastefully redirect users
A good strategy is to tell the user right at the beginning of a conversation that if they get stuck, they can always just ask for help. You’re giving them an escape path right away that is universally understood. When your bot receives a request it doesn’t understand, you can serve up your creative fallback message, followed by a second reminder that help is one input away.
4. How to make a chatbot that excites users?
1. Focus on being conversational.
The way we talk is very different from the way we write, and the way we write and speak is significantly different from the way we message one another. Loaded with colloquialisms, idioms, misspellings and emojis, the language of texting and instant messaging has taken on its own style. This can be one of the biggest hurdles for a bot to overcome when trying to properly understand and respond to a user.
A good way to overcome the hurdle, and improve your conversational design, is to prompt your bot to redirect the user to something it can help them with. For example, a bot could resolve to saying “Sorry, I didn’t get that. Can I try to help you with your search?”, or “I’m better at that”, or “try asking a different way …”. Then displaying buttons for the user to decide if they want to search or ask a question, can help allay any conversational blocks.
This kind of response is significantly more inviting than an error message or a message that simply asks to retry. This would also be a good area to offer an option for human assistance, especially in customer service, in case the plausible conversational flow isn’t in the bot’s scope.
2. Continuously improve based on user feedback
The standard methodology for product improvement is to take feedback from customers, and learn what works best for them Aligning your product’s future with customer preferences is indeed the best way to improve it. Chatbots have an inherent advantage in collecting feedback because the feedback loops can be built into the bot.
Developers need to conceptualize the key success factors for the chatbots they develop and build feedback mechanisms directly into them. The data derived from the feedback can then be enhanced with machine learning tools that can assess results, and iteratively adapt the chatbot to refine its operation.
3. Know your audience.
The difference between an effective chatbot and another annoying message is knowing your audience. It is important to know how to communicate with them, and what their business needs are. Good business chatbots target a specific audience and provide useful information to the user.
The key to development of an effective chatbot is working from data you have about your customers and gaining a better understanding of what they want from your business.
Chatbots are at their most effective when they know what the user is interested in and can apply some level of context to the conversation. This means your bot needs to retain information about the user throughout a conversation, and across sessions.
To make your bot more personable, make sure it can recognize returning users. When the bot needs information (for example, the user’s location in order to recommend a nearby store), it should pull from whatever data has already been supplied and then verify with the user to see if it’s changed. You might also script the bot to begin a conversation by reminding the user of where the last conversation left off.
4. Add empathy to chatbot responses with sentiment analysis
Sentiment analysis is important if you want to improve your chatbot’s conversational design. It helps the bot respond empathetically to frustrated users and prioritize sensitive situations. If a user is especially angry or sad, for example, your bot should switch to more sensitive language and respond to their feelings. Failure to do so will make it sound cold and robotic, and perhaps signal that you don’t understand your users properly.
Bottom line
Chatbots have the potential to be very beneficial to your business and brand if they are created and maintained properly. If you emphasize on human interactions and strive for accuracy, then you’ll be on the right track in creating an effective chatbot for your business. The most important thing to remember is that a chatbot is never complete and there are always scope for improvements. It is a tool that needs to be constantly updated and fine-tuned to be effective.
For more articles around conversational agents, NLP and Deep Learning, follow the dialog
