<img src="http://www.shrfbdg004.com/63997.png" style="display:none;">

Demystifying Artificial Intelligence (AI): Natural Language Processing

Posted by David Lover on Jul 23, 2019 10:00:00 AM

Find me on:

Artificial intelligence (AI) is a hot topic right now. Unfortunately, when a topic is hot, I find that many people throw around the term without really having a good understanding of its details. That’s where our teams within ConvergeOne kick into gear, working to understand those details. As we researched AI, one particularly relevant use case became obvious: chatbots. In fact, Gartner calls chatbots the current face of AI and predicts that by 2020, more than 50% of large enterprises will use product chatbots. Let’s dive into the details of AI within the context of chat.

Chat as a Contact Center Channel

Chat is a critically important channel within the contact center. Not every customer wishes to talk to a human on the phone. Those customers want to gather information themselves, either through a website, Interactive Voice Response (IVR), or some other self-service channel. However, not every customer need can be figured out ahead of time and statically programmed into those tools, so chat has typically been the middle ground. It lets people who are already surfing your website ask specific questions in a comfortable format.

The problem is that chat typically requires a human on the other end, somewhat defeating the benefits of the previously mentioned self-service tools. The opportunity here is to create a chatbot that can talk to the customer in a natural way and answer their questions. That’s where AI comes into play.

Let’s look further into what’s actually happening when AI is used in this scenario of automated chat.

Natural Language Processing

When trying to automate a chat conversation, the very first critical task is to figure out what the user is talking about or asking for. In Natural Language Processing, this is generally called “intent.” As in, what is the intent of the user’s text? There are a million ways to express a given intent. You can’t possibly program a static response for each scenario, so you look for intent.

ConvergeOne has started out by playing around with different technologies, including wit.ai and IBM Watson. One example involves analyzing a variety of questions and requests:

  • “What is the weather in Paris?”
  • “Give me the current weather in Paris”
  • “Is it sunny or rainy in Paris now?”

The very first thing you have to determine is that these questions are all asking about weather, so the categorized intent is “weather.” From there, to respond correctly, you’ll have to look for more information, such as location and possibly time. If you’re missing some of that information (e.g., “What’s the weather like?”), you could make some assumptions about location (the user’s current location) and time (now) and respond with, “The weather in Paris is currently sunny, with a chance of rain this afternoon.” Instead, you could ask for more details: “For which city would you like to know the current weather?”

Now that you know what you have to do, how do you actually do it? How do you program something to be flexible enough to respond correctly when presented with millions of variations? You can’t just do simple searches. It would take too much software writing, looking for specific words like “rain,” “weather,” or combinations of related words.

One of our employees has been diving deep into AI-based Natural Language Processing. He says it reminds him of diagramming sentences, a task we all had to complete during grade school. Remember having to put the subject on the left, the predicate on the right, adjective/adverb modifiers on slanted lines, and so on? Following those old school rules helps get rid of the fluff of a sentence and get to the point—the intent—of the sentence. However, that fluff can add color, such as sentiment or emotion. From this perspective, it makes sense that linguistics is a prerequisite study to get into the field of AI.

Fortunately, there are tools out there that can do this work for you, including identifying entities, keywords, categories, relations, semantic roles, sentiment, and emotion. You no longer have to write these things from scratch, but whether you’re talking about modern contact center technologies, the Internet of Things, or plans for robotic world domination, you will absolutely require a strong understanding of Natural Language Processing. You can count on ConvergeOne to be leading that charge, with the exception of the robotic world domination part! We have strong expertise and products to help you add AI and chatbots to your customer experience strategy.



Automation + Advanced Applications

Want to learn how chatbots can streamline your contact center? Sign up for our proof of concept to realize how this can benefit your business.

Take the first step

Topics: Customer Experience


 

David Lover
David Lover  -- David is a leader in our Office of the CTO and works with every part of the business. From Sales to Professional Services, from senior leadership to end-users, from overall business strategy to nuts and bolts technical understanding, his skills at identifying, articulating, and managing our strategic technology direction to customers, partners, and employees sets ConvergeOne apart as a leader in our industry. David is a former Senior Engineer at Lucent Technologies and Avaya and has applied communications technologies in a business environment for large Fortune 500 and Enterprise multi-site corporations. David is a nationally recognized keynote speaker and presenter at numerous industry conferences, forums, and seminars across the United States. He has built tremendous, strategic relationships with analysts and manufacturers alike, insuring relevancy and the best possible “future state” outcome for ConvergeOne and its customers.