Introduction

There are many factors that contribute to the successful management of an organisation, but among the most important is an understanding of your people. The perspectives of employees on inclusion and equal opportunity, is as important as diversity-related initiatives. Of equal importance, is the collection of customer feedback – between the two, leaders can improve the quality of goods and services.

Traditional methods of relevant data collection include surveys, interviews and social media and online reviews – these allows organisations to collect qualitative insights. Through these disparate channels, employees and customers can express their thoughts, in their own words, which leaders can then use to inform their policy and decision-making, thereby improving their long-term development.

However, the analysis of long-term textual data can be time consuming and, inherent therein, are human biases. This prompts the question – how can we navigate this challenge and analyse open-ended survey responses wisely, and efficiently?

In our paper, “Enhancing Sentiment Analysis of Open-Ended Responses through Advanced Natural Language Processing (NLP) Techniques”, we explore an innovative way of interpreting free text responses. We found that using Natural Language Processing (NLP) techniques can effectively handle complex and ambiguous text and generate a comprehensive, strategic, and consistent report.

To see how NLP can improve sentiment analysis of open-ended responses—and why it matters—read on!

What is Natural Language Processing (NLP)?

There is an unfounded perception that Natural Language Processing (NLP) is yet another complicated analysis tool in a seemingly endless tech world, but if you bear with us here, you might find it’s actually a very efficient AI resource that organisations can use to unlock valuable insights from data.

Briefly, NLP is a form of machine learning technology that reads and understands free-text responses in a meaningful way. Imagine, if you will, having a robot that can quickly read through open-ended feedback, translate it, tag it, and even assess sentiment in mere seconds. Some examples of how NLP can read textual data to make a difference for your organisation:

1. Encourage a culture of psychological safety in the workplace, where employees feel comfortable expressing their views and concerns.
2. Help leaders to identify areas for improvement.
3. Keep on track to maintain high quality of goods and services.

NLP can also contribute to thematic analysis in a few different ways:

• Minimise human biases in categorising feedback
• Identify current trends through keyword clustering
• Enhance inclusivity by discovering workplace culture issues
• Generate comprehensive reports for leaders

Compared to traditional ways of interpreting survey data, employing NLP can serve as a significantly more efficient approach to improving processes and workplace culture.  

Innovating Sentiment Analysis

In our research, we have developed a sentiment analysis system using NLP to understand employee feedback.

Here’s a breakdown of how the system works:

1. Collecting Comments: We start by providing open-ended comments to the system, cleans and formats the data consistently.  
2. Translating in English: The system looks at each comment to detect in which language it is written. If it’s in a language other than English, the NLP translates the comment into English so that native English speakers can understand it.
3. Tagging Comments: We can add our own tags or choose from suggested ones to organise the comments. This ensures that we cover all the important topics mentioned by the employees.
4. Analysing Sentiments: The BERT model, a language model, in the system classifies each response as either positive or negative.
5. Thematic Summarisation: GPT-4o in the system generates summaries of comments related to each tag.
6. Result and Interpretation: The system generates a final report that includes information on the percentage of responses per tag, overall sentiment distribution, and detailed summaries specific to each tag. It also highlights areas for improvement and offers recommendations to your organisation.

By following these steps, we can effectively process textual feedback! There is, however, still room for refinement, as the accuracy of the model is limited compared to human interpretation.

Why This Matters to Leaders: Foreseeing the Trend of NLP

For leaders striving to create a diverse, equitable, and inclusive workplace, gathering feedback from employees is essential. With the rise of “AI for the culture”, many organisations are now using AI to enhance their organisational culture and make informed decisions. That’s where NLP comes into play, offering valuable insights into how things are really going within the organisation. This approach not only helps in addressing employee concerns but also creates a culture that’s open and responsive.

There’s no doubt that AI is set to become a central part of our workplaces. As we move forward, we believe that NLP will be a valuable tool for analysing employee feedback. However, let’s keep in mind that AI is still developing, and there’s always room for improvement!

Here are some key areas to focus on for better sentiment analysis system:

1. Enhanced Contextual Understanding: By training models with specifically labelled data, we can improve how well the system understand nuanced feedback and capture the true intent behind what employees are saying.
2. Bias Mitigation: Addressing biases in data and models is crucial to ensure fairness and inclusivity. This leads to more reliable summarisation and interpretation.  
3. Automated Reporting Efficiency: Developing automated report generation can streamline the analysis process, providing comprehensive insights and visualisations that improve the user experience.

By paying attention on these three areas, leaders can mitigate potential issues associated with NLP and increase the accuracy, reliability and usability of their insights gained from employee feedback. It’s exciting to think that using AI tool for feedback analysis is likely to become a popular trend.