Is the overall energy around open source in your organization translating to individual or team results? If not, consider revisiting your open source strategy.
How to explain natural language processing (NLP) in plain English
What is natural language processing (NLP)? How does it work? Where is NLP used? We break down this branch of artificial intelligence in plain terms so you can explain it – even to non-techies
How does natural language processing work?
One of the secondary beauties of NLP is that it’s arguably the easiest sub-discipline of the broader AI field to explain to non-technical people, especially compared to topics like, say, deep neural networks. The definitions above speak to this. Simply put, NLP is how computers learn to understand and respond to human language, written or verbal.
Let’s revisit the email example above. When your email app suggests words or sentences as you type, it’s not merely guessing. Rather, Burstein from ETS notes that the app is using computational methods to analyze your text (or speech, if you’re speaking to a smartphone assistant or chatbot) and make recommendations as accurately as possible. (Multiple experts note that we’re not all the way there yet on the accuracy front; in terms of sophistication, it’s still early days for NLP, as with other AI disciplines.)
“Behind the scenes, NLP is looking at what you have written to provide suggestions about the next word or next few words,” Burstein says. “This is based on the likelihood of a particular sequence of words. For instance, we are more likely to start an e-mail with ‘How are you?’ than ‘How are your plants growing?’”
In short, NLP does this by converting human language into terms a machine can more readily process. And you don’t need to be a wizard with algorithms to grasp the basics of how it works.
“Under the hood, NLP converts each word or sentence into a list of numbers and applies math to those lists to combine them in different ways and make a decision. This all happens in a split second, ideally,” Manderscheid explains. “Engineers make this work by ‘training’ the technology; they give their computer programs lots of data about language – written sentences and phrases, and transcripts from live conversations – so that it learns over time how words go together, what we’re implying with our words, and what we need from our communication. Fundamentally, it’s not very dissimilar from how babies learn language.”
Natural language processing examples and benefits
Another tactic for explaining NLP to wide audiences is to put in the context of how a business can benefit. It’s not just about comprehension, but meaningful insights. Stephen Blum, CTO, PubNub, summarizes those benefits – which extend well beyond asking Siri for a weather report – like this:
“NLP is a service that can understand the way we communicate – writing, speaking, and even abstract things like art, and derive insights from them. It’s used in chatbots to actually understand what a person is saying. It’s used by brands to keep tabs on massive streams of social media information. And it’s used by businesses to make sense of the massive amounts of data they have at their fingertips, and make data-driven recommendations on it, far beyond what a human analysis could do.”
Havens from Qordoba puts NLP in terms many IT pros can empathize with: Explaining your job to non-technical friends and family. He starts by pointing out that computers are very good at some tasks that are inherently challenging for humans – for example, multiplying large numbers, taking derivatives, checking the value of a sensor 10,000 times per second. (What, you don’t do that for fun on the weekends?)
On the flip side, Havens notes that humans are naturally better-suited for other kinds of tasks, such as walking gracefully, identifying whether two photos have the same people in them, or understanding someone’s meaning when the say something imprecisely. NLP is all about this last scenario.
One way to think of NLP is that it is fundamentally about making it easier for people to interact with machines without needing programming or technical skills.
“Most of the time, when a human wants to talk to a computer, the human has to accommodate the computer’s weaknesses by speaking to it very precisely – in a computer language like Java, [for example]. This is a high barrier to entry,” Havens says. “The goal in NLP is to make the computer do the hard work of understanding what the human meant, instead of making the human do the hard work of communicating precisely.”
[ Want lessons learned on AI? Get the new HBR Analytic Services report, An Executive’s Guide to Real-World AI. ]