Using Natural Language Processing for Sentiment Analysis

The potential of natural language processing (NLP) is significant— the technology can screen resumes, match candidates, analyze employee feedback, assist with performance management and improve harassment awareness.

Understanding employee sentiment has always been important, but this feedback has sometimes been challenging to ferret out in an objective way. New technologies like artificial intelligence, NLP and machine learning are changing this, making it possible for organizations and their HR teams to get data in real time to help them monitor and manage employee sentiment far more proactively than ever before.

“NLP models surface issues that would otherwise fall into a manager’s blind spots—or get lost in a sea of comments—to their top-of-mind priorities,” said Daniel Norwood, vice president of marketing at Perceptyx, an employee experience platform based in Temecula, Calif.

But what exactly is NLP? How does it work in HR? What are its pros, cons and best practices?

What It Is and How It’s Used

“Natural language processing and machine learning techniques help organizations determine whether employees express positive, negative or neutral opinions in their survey answers,” Norwood said. “It also helps companies effectively determine common themes emerging from employees. Specifically, sentiment analysis allows companies to get deeper insights into survey responses, allowing them to turn their insights into actions.”

For example, NLP can be used to gauge employee sentiment following training sessions to identify areas of opportunity for improvement. During the talent acquisition process, it can be used to identify potential “red flag” issues in applicant materials. From an employee engagement standpoint, it can be used to assess employee buy-in on topics related to company culture and strategy.

“Advancements in AI with NLP models can analyze hundreds of thousands of employee comments to determine intent, theme and sentiment,” Norwood said. “Managers find this extremely valuable to sift through vast amounts of employee feedback very quickly. These models can effectively identify the different concerns and allow the manager or HR to respond appropriately.”

There are a wide range of applications for NLP in HR. Norwood said, “We often see customers use NLP to better understand employee feedback expressed through surveys, digital focus groups and other listening events. This can be especially helpful when the volume of this input is significant, as in very large enterprises.”

He said NLP models can be used to flag sentiment and themes or categorize feedback according to emotion or intent.

Tim Glowa is CEO and founder of HRbrain.ai, an AI-powered bias detection tool. Glowa said that sentiment analysis is used at HRbrain.ai not only to detect potential biases in communication but also to assess corporate culture.

“By employing NLP, we offer companies insights into the nuanced dynamics of their workplace environments, enabling them to understand and improve employee sentiment effectively,” he said. It’s an approach that can be “particularly impactful for identifying systemic issues within teams and across the organization.”

Potential Drawbacks

Sean Spittle, a software developer and software-as-a-service expert, said that when the HR department at InspectNTrack, safety inspection software based in Hinckley, Ohio, first proposed incorporating NLP sentiment trackers, “visions of innovative metrics and streamlined decision-making danced in all of our heads.”

The reality, though, “has been more complicated,” he said.

On the upside, Spittle said, “our NLP algorithms have unlocked fascinating insights from open-ended internal surveys and feedback forms that manual analysis would have likely missed.” By parsing textual data from emails and messaging channels, the company “can now quantify previously invisible trends in morale, collaboration, managerial issues and more on a companywide level.”

The power to make data-driven talent management decisions “is undoubtedly rewarding,” he said.

However, he acknowledged, wrestling with biases in the NLP models has proven challenging. “Without extremely thoughtful dataset curation and indicator selection, these tools risk misinterpreting sarcasm for sincerity or amplifying problematic assumptions. Our legal counsel also rightfully flagged the dangers of infringing on employee privacy, informed consent or labor rights if implementation feels coercive rather than collaborative,” Spittle said.

NLP certainly holds promise for HR leaders and their teams but, like any technology, its efficacy is dependent on the humans that use its outputs. There are also certain aspects of its purpose and use that can raise concerns among employees, specifically over privacy and security.

Ensuring privacy and transparency are critical, said Daniel Wolken, an HR expert and talent acquisition specialist at DailyRemote, a remote jobs board based in New York City. But, he said, “with the right ethical safeguards in place, NLP can positively transform how you tap into employee sentiment. As with any new technology, change needs to be handled thoughtfully,” he said. “Ensuring transparency around how information is collected and used will be key to gaining employee trust.”

The potential for bias also must be guarded against, Wolken cautioned. “The last thing we want is for certain groups to be unfairly impacted.”

Best Practices

Conor Hughes, SHRM-SCP, an HR professional and consultant who shares his expertise on SMB Guide, a buying resource for small and midsize businesses, recommended starting with a limited-scope pilot “focused on language patterns most relevant to your goals.” Even basic techniques, he said, “can uncover impactful insights to improve company culture, retention and performance.”

He pointed to a client who used NLP and discovered a high prevalence of anxiety and uncertainty in emails from one of the company’s divisions. This allowed the company to proactively address underlying issues through training and communication.

Hughes said that he has seen clients “uncover problems before they escalate, improve managers’ emotional intelligence and create more meaningful engagement surveys.”

Based on Spittle’s experiences at InspectNTrack, he said, “I now question whether some snippets of language are meant to be codified, even if the capabilities exist.”

His advice to others: “Carefully examine if understanding an emotion will necessarily help address its underlying causes. NLP should illuminate shared truths, not impose top-down rules. Maximizing profit should never justify minimizing dignity. And if these tools inadvertently neglect minority voices or concerns due to dataset limitations, they have no place making high-stakes HR decisions.”

Used in the right way, Spittle said, “NLP-fueled sentiment analysis tools can indeed positively transform workplace culture and productivity.”

However, it’s important to exercise patience, explore ethical issues and call upon human emotional intelligence that algorithms just can’t replace when evaluating outputs.

Technology exists today to help HR leaders in a variety of ways. It’s important, though, to consider the potential for risk as well as reward, to learn from others’ experiences, and to approach potential use cases on a small scale to determine any issues before rolling tools out more broadly.

Lin Grensing-Pophal is a freelance writer in Chippewa Falls, Wis.