The potential uses for Sentiment Analysis are limitless: a historian can use sentiment analysis to understand the intent of an author writing hundreds of years in the past. Likewise, a marketing manager can monitor the evolution of brand reputation over time.
The Sentiment Analysis method discussed in this article will use machine learning to score your text and classify it as expressing Positive, Negative, or Neutral emotions.
You will need Microsoft Excel and the Azure Machine Learning Add-in.
Why Is Sentiment Analysis Important?
For people who build products, work in marketing or politics, or are conducting research, understanding the emotional sentiment regarding a particular subject is a professional necessity.
Sentiment Analysis can help them. While it will not entirely replace usage data, surveys, interviews, and desktop research, Sentiment Analysis is a solid tool to have at your disposal.
Why? In almost any situation where you have a large amount of unstructured qualitative data, sentiment analysis can quickly give you insights into its underlying message.
Sentiment analysis works best when a large amount of data is analyzed.
Performing Sentiment Analysis on the most recent text message from your romantic interest is unlikely to return information with any added value. On the other hand, analyzing thousands of Tweets containing a specific hashtag will give you useful results.
Other possible use cases include analyzing product reviews, reviewing customer surveys, and uncovering a public relations crisis. In addition, regular sentiment analysis will allow you to track how customer attitudes towards your company are changing over time.
Volume vs. Sentiment
Sentiment Analysis is an essential part of social media monitoring for any company or brand conscious of their reputation.
For example, you may see that your company gets a large volume of mentions on social media. But mentions alone aren’t everything.
Sometimes mentions are a good thing. For example, they can mean a large amount of positive public sentiment towards your company.
Other times, you may be facing a PR crisis that is spiraling out of control. As a result, the public sentiment towards your company is negative.
Distinguishing the sentiment within a high volume of social media mentions can make all the difference.
Using Microsoft Excel for Sentiment Analysis
Some social media monitoring platforms include sentiment analysis as part of their offering. It is also possible to perform sentiment analysis on text using a programming language such as Python.
However, these options require either a significant budget to afford a social media monitoring platform or coding skills.
If you’re like most people and don’t have either of these, Microsoft Excel is a good option for performing fundamental Sentiment Analysis.
While none of these tools produce perfect results, they can help you understand the overall trend of the sentiment contained within the text.
How to Perform Sentiment Analysis in Microsoft Excel
Follow these steps to try out a sentiment analysis with Excel without writing code. Under the hood, Excel and the Azure add-in depend on a natural language processing algorithm and a generic dictionary with positive and negative words. Each word in the lexicon is assigned a positive, neutral, or negative value.
Organize the data you want to analyze in a Microsoft Excel Sheet. Clean up the data by removing blank spaces and unnecessary characters. Make the first cell in your dataset tweet_text (keep in lower case). Go to Insert > Add-ins. Next, head to Search > Azure Machine Learning. Once installed, the Azure Machine Learning add-in will pop up a box on the right side of your screen. You’ll see two options: Titanic Survivor Predictor and Text Sentiment Analysis. Click on Text Sentiment Analysis. Go to Predict > Input, then add the range where the data you want to analyze is located. Leave My data has headers checked. Go to Output and add the cell where you want the analysis results to go. Press Predict.
A Sentiment and Score for the text in each cell will populate; the corresponding text is more Negative if the score is closer to zero. You may prefer to change the Scores to a Percent. In that case, the closer a Score is to 100%, the more positive it is. Neutral is any Score around 50%.
See the below example from Treasure Island by Robert Louis Stevenson.
How to Obtain Insights From Sentiment Analysis
After running the Sentiment Analysis, you will have cells with Positive, Negative, or Neutral classifications and their corresponding numerical scores.
How can you turn the results into understandable insights? Here are a few ideas:
Segment the classifications by creating a Pivot Table in Excel. You can use Visio, which is now included in Microsoft 365 Business at no extra cost, to visualize the overall number of each of the Positives, Negatives, or Neutrals. Data visualization can give you a bird’s eye view. If you are responsible for reputation management at a company or brand, you may want to focus on scanning through all the texts classified as Negative. What makes the text Negative? Is there something you need to pass on to address the issue? You can do the same exercise for the texts classified as Positive. Maybe there is a particularly nice customer testimonial buried in a high number of product reviews that you would like to share. You could also further segment the text, so you only see cells that mention a new product feature. Are users more Positive, Negative, or Neutral about the feature? Sentiment Analysis can help you determine this and more efficiently gather feedback.
Sentiment Analysis can take people out of the decision-making process. Sometimes this can be good because text interpretation can be highly subjective.
Use of Microsoft Excel for Sentiment Analysis
If you want to try performing Sentiment Analysis but don’t have a lot of financial resources or coding skills, then Microsoft Excel is an excellent place to start.
Sentiment Analysis in Microsoft Excel will give you insights that you can use to understand unstructured text data. It could also be an ideal way to familiarize yourself with machine learning concepts before diving into a project in the field.
