<img height="1" width="1" alt="" style="display:none;" src="//www.bizographics.com/collect/?pid=5286&amp;fmt=gif">

Digital Engagement Blog

Understanding NLU - A Cheat Sheet for Beginners

NLU is a hot topic these days as it it powering many conversational interfaces. 

So I asked our speech experts to help explain NLU.

But first to Breaking Headlines:

  • Iraqi Head Seeks Arms
  • Ban on Nude Dancing on Governor’s Desk
  • Juvenile Court to Try Shooting Defendant
  • Teacher Strikes Idle Kids
  • Stolen Painting Found by Tree
  • Kids Make Nutritious Snacks
  • Local HS Dropouts Cut in Half
  • Hospitals Are Sued by 7 Foot Doctors


These headlines show that language is complex, ambiguous, flexible, and subtle.

A contact center agent can understand the headlines but does a machine/computer/IVR/chatbot have any hope of understanding what is said?

NLU to the rescue!

One technology that can help is NLU.  It can understand the meaning of conversations, dialogs and spoken interactions. 

If it's not already, NLU should be in your customer engagement strategy toolbox.

In this NLU Cheat Sheet, we cover definitions and basics for NLU.

What does NLU stand for? 

Natural Language Understanding

What does a NLU do? 

NLU understands a customer's intent.

Its a technology that lets users interact with a system using natural language and an experience around it design to simulate understanding.  Like talking to a good friend.

NLU Understands

While humans are able to effortlessly handle mispronunciations, swapped words, contractions, colloquialisms, and other quirks, machines are less adept at handling unpredictable inputs

Human conversation is complex, so machine understanding of spoken words is  complex.

Give an NLU engine the text of a customer talking,  and it is designed to understand what the customer is saying.  NLU determines the customer's intent. 

If NLU cannot understand the customer, then it can ask questions to determine caller intent.  This process is called disambiguation.

NLU is a complex system that uses big data and machine learning to do text mining and predicitve analytics.  It extracts meaning from text.  We're not going into the technical details here,  but we can give you a general idea of what NLU does and does not do.   

What does NLU not do?

NLU is not an out-of-box solution that “just works.” NLU solutions are typically domain-specific, with intensive data gathering, human intervention, and subject matter expertise needed to train the system.

Where is NLU used? 

As NLU becomes widely available, more products & services embed NLU into their offerings.  Here are 11 different uses:

  1. natural language IVR (Contact Solutions, A Verintcompany)
  2. voice-driven assistants (Siri, Google Now, Microsoft Cortana)
  3. chatbots (Watson chatbots, lots more)
  4. natural-language search (Google, Facebook Graph Search)
  5. question answering (Google, IBM’s Watson, Wolfram Alpha)
  6. web-scale relation extraction (Google, IBM DeepQA, startups)
  7. sentiment analysis for automated trading (many hedge funds, Sentiment Alpha)
  8. legal discovery (Cataphora, H5, IBM)
  9. business intelligence (Palantir, Quid)
  10. social media analytics (a zillion startups)
  11. content summarization (Agolo, others)

NLU router

In a speech IVR system, an NLU router's function is to understand the customer’s intent, manage the conversation, personalize the interaction and route to the best next action.  The router allows customers to use normal spoken language to describe their issue instead of forcing them to navigate a complex menu structure. 

Get the Experience Right

Important point:  Get the experience right

Most speech IVR offerings (and many other systems) get NLU wrong: it’s only one piece of the pie.

To get NLU right you need more.

NLU needs to be wrapped with a well designed experience.  A design that takes into account-

  • goals of the business
  • agent queues
  • self-service and automation options
  • ways the customer can ask a question
  • context
  • customer profile and history
  • and more

When you understand these aspects you can  build an NLU system with the right level of granularity necessary to get the experience right.

Having an NLU system understand 250 distinct meanings is only useful when you have deep user experience designed around each and every one of those meanings.


Once you understand the caller's intent, you can design a very good experience that reduces effort and friction-- one that places the burden on the system, not the user.  Here is an example experience flow-

  • a customer calls into a care center
  • immediate routing of a call to an agent
  • forward to another IVR system
  • play a variable message
  • perform an outage check
  • send a signal to refresh equipment in the home
  • offer an upsell
  • read out account information
  • verify the user’s identity, and so on. 


NLU is powerful technology.

But we recommend you consider the end-to-end customer experience surrounding the technology to be equally as important as the technology.

Design matters.

Forrester Research Study

Exclusive Forrester Research Study

New insights on how to jack up Agent & Customer Experience with IVR+WFO analytics.

Get the Study >

Topics: Performance Optimization voice self-service User Experience