I made some predictions about the future of chatbot a few weeks ago; many users came to ask me about the differences between NLP (Natural Language Processing) and CI (Conversational Interfaces). Today, I would like to take a close look at these two different approaches.
NLP, the Terminator Approach of Chatbot
First, let’s define what NLP is: it stands for Natural Language Processing, it’s a type of artificial intelligence technology that aims to interpret, to recognize, and understand user request in the form of free language. When most people first think about chatbot they would imagine some super intelligent computer like the Terminator.
However, the reality is that the NLP technologies are still in very early stage and the average precision rate is around 60–70% which is not very usable for certain cases.
Some good examples of NLP based chatbots:
x.ai
X.ai is an email-based personal assistant, who schedules meetings for you. Founded in 2014, the team of x.ai use their knowledge of users’ schedules and availability to respond to incoming scheduling requests. The company’s AI analyzes return responses and, then, automatically sends out calendar invitations.
Xiaoice
Xiaoice is an advanced natural language chat-bot developed by Microsoft. It targets the Chinese community on the microblogging service, Weibo, primarily. The conversation is text-based. The system learns about the user and provides natural language conversation. Microsoft gave Xiaoice a compelling personality and sense of “intelligence” by systematically mining the Chinese Internet for human discussions.
Mitsuku
Mitsuku, is a chatbot was created from AIML technology by Steve Worswick. Mitsuku won the 2013 and 2016 Loebner prize. It is available as a flash game on Mousebreaker Games and Kik Messenger under the username “Pandorabots” and was available on Skype under the same name but was removed by its developer.
CI, the rising star of chatbot
On the other hand, we have CI, which stands for conversational interfaces. I might be biased, but I think this is the best approach that will take chatbot to the mainstream market. I wrote some time ago a piece explaining this concept.
A simple definition would be a CI is a hybrid UI that interacts with users, combining chat, voice, or any other natural language interface with graphical UI elements like buttons, images, menus, videos, etc.
Unlike NLP which focuses on understanding what the users say, CI focuses on providing what users need: a personalized experience.
Some Examples of Conversational Interfaces
Landbot.io:
It is a saas platform that aims to humanize website experiences with conversations. Landbot help companies to engage more with their user and boost their conversion rate significantly. To demonstrate the potential of conversational interfaces, the Landbot team has built its whole website in a conversational format.
Adrian Zumbrunnen:
Adrian is a UX/UI designer who has transformed his personal website into a conversation. He also wrote some interesting learnings about his experiment.
Upcoming pages:
Product Hunt, the premium platform for launching new digital products, has recently released a new feature called Upcoming Pages. It helps product makers to get early users for their upcoming product.
Which one is better?
Now we know the difference between NLP and CI, but which one is better? In my opinion, both technologies have their pros and cons, so we should not ask “which” is better but “when” is one approach better.
A tool can be good or bad depending on the context in which you use it. So, let us go through various scenarios, and see how it plays out.
Companionship chatbot:
This a particular type of chatbot because its sole focus is on keeping the conversation with a user open when he or she feels lonely, stressed out, etc. For many people, the need for a companionship bot might be hard to understand. However, there are many parts of the world where people are under high social stress, which makes this kind of chatbot a good alternative.
As the main goal is to keep the conversation going and let the user feel like talking to someone who understands his pain, the best approach is NLP. The chatbot doesn’t need to have a high precision rate of understanding the user’s language but their emotional feeling. As human sentiments are a highly abstract concept, it allows the chatbot to provide general responses to the user but still make sense to continue the conversation. A very good example is what Xiaoice mentioned before.
Website chatbot:
As I have argued in the past, it is one of the killer use cases for conversational interface. I already explained the reasons here; some key points are:
CI by nature focuses better on the user attention, which helps to drive engagement. The potential of CI is to personalize each user interaction and always offer the most relevant information. Being able to qualify leads in real time is a huge benefit for companies to optimize their full funnel conversion.
To ilustrate this use case, one great example is Landbot.io
Customer support bot:
One of the first use cases many people saw about chatbot is using it to replace customer service agents and automate support tickets. The principal characteristics are the urgency and the specificity of chatbot that users have when they interact with a customer service chatbot. In this case, the ideal solution would be based on NLP. However, as most NLPs are lacking a good precision rate, some human takeover capabilities are needed to solve complex cases where the NLP fails.
As rule of thumb, if we look at the customer journey as a funnel, the later stages the users are in, the more need is there for NLP solutions.
Hybrid is the king
As we can see in the previous examples, each approach has a better fit depending on the context of our users. The reality is that most chatbots in the market are operating with a hybrid model, where they combine the use of NLP and CI to offer the best UX possible.
Some examples are as follows:
Google Assistant, the smart AI bot which combines voice input (NLP) and CI to help people in different tasks. Previously only available in Nexus and other direct Google brand smartphones, but a few months ago, Google made a significant update to make GA widely available for every Android phone. Bellow, you can see some interactions from GA.
Lidl´s wine concierge on Messenger. In this case, Lidl builds a simple chatbot that can understand some basic user request (NLP) regarding wine selections and then offer different recommendations to the user in the form of images (CI). If they click the image, the bot will redirect users to the payment cart to buy the wine.
Alexa Echo Show. It’s one of the few Echo devices that has a visual screen. The interaction with Echo Show is a hybrid; you can make voice commands (NLP) but also use the touch screen (CI). According to Alexa VP: Voice won’t kill the graphical user interface https://venturebeat.com/2017/10/24/alexa-vp-voice-wont-kill-the-graphical-user-interface. He argued that for specific use cases such as exploring a list of products, it is better to have a visual interface than to be based purely on voice.
Conclusion
To conclude, I would like to share my personal view about NLP vs. CI. In my opinion:
The future of chatbot is not about building better natural language understanding but about designing smarter conversational experiences.
Your customers don’t care about if they can use natural language to interact with computer, as long as it’s simple and easy it would be enough. What they really care about is to get a frictionless experience after they made their request. The Chatbot community should shift the focus to think how to provide a more personalized experience to the customers.
Did you know you can give up to 50 claps by continuing to press the hands? Please give 👏 ‘s if you enjoyed reading it means a lot!
For more content from me please subscribe to my WhatsApp Newsletter 🙌
