We regularly talk about employee engagement being important. It’s good for employees because it makes them feel valued and connected to their work. It’s good for business because engaged employees are more productive and will stay. But trying to find out if employees are engaged can be hard.
It’s one of the reasons that organizations conduct surveys. Hopefully, the survey results will tell us what employees are thinking. The reason I say “hopefully” is because there are times when we let our human biases read too much (or not enough) into employee feedback. Ideally, organizations need an unbiased way to interpret survey results.
That’s where sentiment analysis comes in. Sentiment analysis is a technology that helps organizations understand employee comments. I had an opportunity to learn more about this during a demo of Workhuman’s Moodtracker product and asked Yi Chu, natural language processing manager at Workhuman if she would share some insights. Thankfully, she said yes.
What is sentiment analysis? To be specific, is it a form of Artificial Intelligence (AI) or is it more accurate to say it utilizes AI?
[Chu] It utilizes AI. Sentiment analysis is an application of natural language processing (NLP) techniques. NLP is an area of AI. Simply put, sentiment analysis uses NLP and machine learning techniques to analyze large volume of text feedback to detect positive or negative feelings and uncover underlying opinions.
Obviously, HR pros don’t want to get a degree in engineering to understand sentiment analysis and sell it to the organization. Can you share with us a workplace example of how sentiment analysis works?
Yi Chu Workhuman headshot
[Chu] One example is that sentiment analysis can be applied to open-ended feedback of employee surveys to rapidly gauge the employee’s perception towards the organization and how they feel about work.
More advanced techniques, such as aspect-based sentiment analysis, can be further applied to pinpoint the topics or themes that are mentioned in the surveys to give a better understanding of the survey responses.
Many tools that use sentiment analysis will share the results in layman’s terms, so the results don’t require an analytics or scientist to understand the outcomes.
Since sentiment analysis helps organizations achieve greater insights, how does this differ from and complement existing feedback tools like employee surveys?
[Chu] Sentiment analysis complements existing feedback tools, it’s not a stand-alone solution. Analyzing qualitative data (i.e., open-ended responses) manually is extremely time consuming and subject to human bias. Utilizing sentiment analysis provide a rapid and systemic way to analyze the data automatically. The analysis is consistent, real-time and easily scalable to large volume of data over time.
To properly understand others in our everyday face-to-face conversations, we have to “read” the other person’s body language, tone, etc. How does sentiment analysis take into account things like culture, slang, etc.?
[Chu] Sentiment analysis looks at text or verbal data so it will not take nonverbal communications into consideration. Yet sentiment analysis can be customized to take into account culture-specific language use or slang by adding the required into the model.
Generally speaking, you need a way to decode the context elements from your data and factor that into the modeling process. Using a survey as an example, you might want to add screening questions to collect the regions of the participants because regions and cultures are strongly correlated. Then region would be used as a variable in your sentiment analysis models when analyzing the survey results.
For organizations considering sentiment analysis, what are 2-3 things they should keep in mind?
[Chu] At the end of day, current machine learning algorithms are not perfect and struggle with interpreting the complexity of human language such as sarcasm, jargon and mixed opinions. Companies that are thinking about using sentiment analysis will want to do these things:
Design open-ended questions in a different way in order to make sure you get genuine feedback from employees. Open-ended are better at soliciting genuine feelings from survey participants than structured questions.
Use the result of sentiment analysis as a reference instead of a dominating factor. Use the results but don’t let them be the ‘end all, be all’.
Run surveys more frequently and apply sentiment analysis on a large pool of employees, since large data sets provide a more accurate and less biased analysis. While there is no standard on how often to run surveys, organizations might want to avoid annual or one-shot surveys because having a regular cadence of surveys scheduled is beneficial.
I want to extend a huge thank you to Yi for helping us understand sentiment analysis. If you want to learn more, be sure to follow the Workhuman blog for more insights. Or watch a demo of Moodtracker to see how sentiment analysis can be used in a workplace setting.
Right now, organizations are trying to figure out the best way to deliver a human experience to their employees. Because they know that creating a more human work environment is good for the employee, the customer, and the bottom-line. Tools like sentiment analysis can help organizations understand what employees want and need.
Source : https://www.humanresourcestoday.com/?open-article-id=15422414&article-title=everything-hr-needs-to-know-about-sentiment-analysis&blog-domain=hrbartender.com&blog-title=hr-bartender