bar graphs, line graphs on dark background

Natural language processing and the power of contextual data

by George Dearing

"Technology in the workplace is as much about power and control as it is about productivity and efficiency."

This is according to Zeynep Tufekci, author of an interesting article in the New York Times¹ about the relationship between power and technology in the workplace. And “power” is the operative word. Tufekci argues that we should be focusing on how we value each other, rather than optimizing the latest algorithm to be a job creator or killer.

The machines-versus-humans narrative isn’t going away though, and to be sure, certain jobs will go away. Various routine tasks are already disappearing in every industry, from journalism to healthcare and manufacturing. Yes, the routine stuff (see Wall Street Journal chart below) is getting eaten by software.

A big enabler in this process is natural language processing (NLP), a process which helps non-programmers get useful information from all those machines. So how is NLP changing how companies operate?

​Describing a company as 'lean' describes how they handle their processes. Employing a lean methodology puts all those Six Sigma and process gurus to work by refining and breaking down bottlenecks into more manageable parts, eliminating waste and driving efficiency.

Leaner organizations with bigger data appetites

Describing a company as “lean” describes how they handle their processes. Employing a “lean” methodology puts all those Six Sigma and process gurus to work by refining and breaking down bottlenecks into more manageable parts, eliminating waste and driving efficiency. NLP is helping organizations rely on fewer people to come up with such answers.

“Every time people come to me and ask for new bodies, it turns out so much of that can be answered by asking the right questions of our data and getting that in front of the decision-makers,” says James Reinhart, CEO of thredUP, speaking to the Wall Street Journal². “I think frankly it’s eliminated four to five people who would [otherwise] pull data and crunch it.”

The knee-jerk reaction to that is: “See I told you, the machines win again!” But not so fast.

All those companies the Journal spoke to “had middle and even senior managers who operated as player-coaches, tasked with both doing things and directing others.” That’s the distinction. There’s still a need for delegation and talented people to decipher meaning from machines.

At least for the time being.

Six questions for NLP projects

A good way to start thinking about NLP use cases is to frame what you’re trying to accomplish. Data marketplace and aggregator Kaggle³ challenges data experts to answer six questions.

1. What problem are you addressing? Who will benefit?

There’s a good chance this is much broader than you think. Spend time mapping your stakeholders. Once you get a good sense of what motivates a particular group, you can start to think about data in a more tailored way.

2. What’s the current solution?

This is a chance to dig deeper and see the fruits of your research. Look for data sources or dormant processes that might be re-invigorated with a fresh set of eyes and improved access to key data. Think about how the use cases might change as data are manipulated and compiled in different ways with newer data sets. That’s the sort of process that spawns new applications and new business models.

Is NLP changing the way you operate?

3. What are the data inputs and where do they come from?

This used to be straightforward and much more limited — but not anymore. With advancements in open data and the push for more transparency, don’t limit your inputs to what’s behind your firewall. Mashups have always been cool, it’s just the terminology that’s changed on us.

4. Will it benefit businesses?

Thinking through a business case will provide clarity here. If you can’t come up with something compelling as to why a business unit should re-engineer a process or how it can surface new opportunities, move on.

5. What are the outputs and how is the information consumed or processed?

Think about how the results need to be rendered. Will it be a static report or will an online algorithm accomplish the analysis and present results?

6. Does it cut costs or increase revenue?

Referback to your business case. There are plenty of business improvementprojects, but the ones that get prioritized have numbers attached to them.

Content’s seat on the throne has been safe for a while. What might be their heir apparent is context. Alissa Lorentz described the scenario in WIRED⁴:

“Take the example of a company that has invested heavily in business intelligence (BI) software that organizes internal data. In an increasingly connected world, this company has not leveraged its data to its potential. Why not? Because the company’s internal data is isolated from the rest of the data universe including news, social media, blogs, and other relevant sources. What if you could join the dots between all these data sources and surface hidden connections?”

That’s exactly where we are in terms of NLP. Many challenges with just as many opportunities.

George Dearing has more than 15 years of experience helping organizations understand how information, technology, and the Internet impact business. As founder of the Dearing Group, he advises clients on strategy, business development and communications. After working for one of the first Internet consulting firms (USWeb) in North America, he’s run marketing groups at software companies, directed strategic alliances at professional services firms, and helped early-stage companies deliver software-based business solutions.

  1. 1Tufekci, Zeypep. "The Machines Are Coming." nytimes.com. 18 April 2015.
  2. 2Mims, Christopher. "Data is the New Middle Manager." wsj.com. 19 April 2015.
  3. 3kaggle.com
  4. 4Lorentz, Alissa. "With Big Data, Context is a Big Issue." wired.com.