Blog & News

Natural Language Generation: More Than a Chatbot

By John Alexander | January 6, 2022
Blog Image - 08 Natural Language Generation More Than a Chatbot

Chatbots have become a standard part of customer service and marketing for many businesses, to mixed reviews from customers. Like all artificial intelligence (AI) solutions, there are missteps in the beginning, but eventually, the technology becomes smarter. At Arria, we work with natural language generation (NLG), a subset of AI that lends itself to smarter conversational interaction rather than rote, programmed answers to presumed questions. 

The NLG that Arria uses in its chat-enabled platform is based on proactive conversational AI. It delivers answers that are context-based and anticipatory. When a system knows the context of the conversation you are having with your virtual digital assistant (VDA), it can correctly identify and anticipate your real intents. By contrast, if the system operates under a reactive model as chatbots do, it treats the digital assistant like a basic query system. 

Conversational AI 

When you think of conversational AI, you likely think of chatbots. Chatbots are often beloved by business for the quick customer service response they allow for, and often detested by customers. They allow the business to be responsive at a lower cost than a human helpline, and sometimes they work well for customers. Often, however, they can only deliver surface-level information, too often not the information the customer is seeking.  

But what about conversational AI that delivers an altogether different experience and is more than a first-line customer service solution? What about conversational AI that allows you to engage with your data more easily and more quickly get answers to the business questions you have? Say you want that. 

And if you want that, do you also want it to employ proactive conversational AI? What does that even mean?  

Engaging deeper conversations 

To understand the difference between the two and why natural language generation (NLG) is superior to chatbot technology, let’s look at what chatbot technology is.  

Like NLG, chatbots are built on artificial intelligence technology, allowing the tech to “chat” with you. A chatbot can answer your questions, to a degree, and provide information based on preset answers to questions the chatbot expects you to ask. For example, voice-enabled services such as Amazon’s Alexa is task-driven; questions or commands such as “Turn off lights” are not a conversation. There are also chatbots designed to help you with customer support, but they are not insightful. 

Think of the difference in language evolution as a child grows. The conversation you have with a 4-year-old is usually to deliver information to the child. The questions they ask are typically easy to answer, and you have the answers at the ready. 

Now, think of the conversation you may have with an adult, a hyper-intelligent person, possibly a data scientist or mathematician. The questions they ask may leave you searching for answers, or, unable to answer because the conversation is anything but predictable. Consider how frustrating it is when a chatbot delivers a “predictable” answer to a question you did not ask. 

What natural language generation brings to the conversation is another level of human interaction and adeptness; artificial intelligence erases much of the frustration chatbots have become known for. NLG generates answers to your questions by connecting ideas and layering in additional information. NLG makes you think harder, think about something else, something beyond your question. 

When it comes to Arria, conversational AI that is proactive means a system that does not just answer your question, it answers the next question you are going to ask too. 

Examples of Proactive Conversational AI 

What does the move from chatbots to proactive conversational AI look like in business? Say you ask your VDA (Virtual Digital Assistant) “How are my sales?” The VDA might tell you “Sales are $1.57 million.” There. You have been given a simple fact. No context, no causes. 

When the answer goes further, to anticipate follow-up questions and answer those too, then you are getting proactive conversational AI: “Sales are $1.57 million, in line with your target” or “Sales are $1.2 million, a decrease of $1.3 million from the same period last year.” When conversational AI goes further, it helps guide you through your data. Conversational AI that is proactive can identify the insights and highlight potential problems or wins for you, so you don’t have to go digging around looking. 

Arria’s focus in the sphere of conversational AI is building proactive systems. Our conversational-AI offering, called Arria Answers, is a platform that highlights key insights without the user having to repeatedly drill down. It can be integrated with any VDA, giving companies real-time access to business insights – delivered using natural, spoken language. 

“Your sales are $1.57 million. Would you like to know why?” 

When the VDA proactively solicits input, it can provide insights more quickly, it reduces ambiguity, and it can lead to better-informed decision-making. Proactive conversational AI gives you confidence that performance indicators are on track without having to query your data over and over. 

What does this mean for you, as an Arria customer? 

With Arria Answers, we are bringing to VDAs our foundational experience in natural language generation – along with our innovative insight analytics. Arria Answers offers machines the power of natural human language, empowering a digital assistant to be smart, contextual, and accessible. 

Smart. Arria Answers anticipates what you want to know. It enables your VDA to answer your next question before you even ask it. A VDA needs domain knowledge to deliver key insights and metrics without reporting every little fact that might be uninteresting or unimportant. It needs to remember and understand your preferences as an individual rather than responding in the same way to each user. And it needs to understand synonyms, idioms, slang, acronyms, and other complex language features that allow you to interact in a truly human way. 

Contextual. Arria Answers remembers where you are in the conversation – and where you are in your data – so you can ask follow-up questions if you want. If you misspeak or if you’re misunderstood, you can re-word part of your question to the VDA without having to repeat everything. Also, since Arria Answers knows the context of your question, you need only specify the relevant part of a request, with Arria Answers filling in the unspecified elements automatically.  

Convenient. Arria Answers encompasses convenience in two ways.  

    • Taking simple natural-language commands, Arria Answers allows non-technical users to interact with data wherever it is convenient or whenever it needs to be hands-free (for example, in the board room, while you are driving, or in a hospital, lab, or warehouse setting). 
    • Arria Answers is easy to use: its ad hoc conversational format allows users to freely interrogate their data without the help of experts. 

How do you turn reactive answers into proactive ones? 

“Sales are $1.57 million.” To get to the context, to get to the causes, you would have to ask more questions to find out what is going on. 

Arria knows what you want to know – “Are my sales as I expect them to be? If not, why?” Because we know this, we can build it into our conversational AI. 

How does Arria Answers know what you really want to know? 

Arria’s analytics algorithms do the heavy lifting and find actionable insights and important metrics, without you having to specify and look for those insights. Arria Answers encodes the semantics of your domain within the product. This means that Arria Answers is smart enough to know which characteristics make an insight important to you and report that insight in a timely and actionable way. 

What kinds of analysis does Arria Answers perform? 

Because Arria Answers is backed by Arria NLG Studio, it can deliver any kind of analysis that you want. The true power of NLG Studio lies in its capability to give voice to your data – your voice, in your way of reporting, with the analyses you choose to apply. But out-of-the-box, Arria Answers delivers variance and key performance analysis. You can effectively see which business units are contributing toward your goals, without having to sift through every one manually. You can see what is driving period-on-period change of key measures, without having to list each potential driver or offset to see if it has an impact within the period. 

How does Arria Answers relate to Arria’s other products? 

The Arria Answers platform works seamlessly with the rest of Arria’s offerings and is easy to integrate as an existing NLG Studio customer. Arria Answers just needs to connect to your data source. 

All the analytics and narratives used by Arria Answers are generated in NLG Studio. What this means is that you can fully customize any response. Say you want a shorter response, or maybe more detail, or even a specific phrasing – with the Arria Answers/NLG Studio pairing, you can have control over the output narrative at every level of detail. 

Want to learn more? 

To get the latest news and find out about advances in NLG, sign up for our blog. 


Dr. John Alexander is Senior Data Scientist at Arria NLG and Product Owner for Arria Answers.