Parlor trick or worthwhile?

1. Parlor trick or worthwhile?

Welcome back. In chapter 3 you will start to make visualizations based on sentiment analysis. In many organizations, part of a data science project’s success relies on a narrative or readout. As a result there are times when a good visualization may aid in educating decision makers.

2. Interesting visuals

When doing sentiment analysis I have two pieces of advice. First, avoid using a lot of word clouds. People love them but they are cliché. For the record you will make a wordcloud or two in this chapter! I much prefer using bar charts which can be just as informative for showing frequencies. Second, make sure you avoid sentiment analysis that doesn’t bear any insight. For example, in one of my leadership roles a potential vendor pitched sentiment analysis for customer satisfaction surveys. This vendor pointed out the surveys with poor polarity scores. The survey itself had a simple yes or no question as to whether or not the experience was positive. So why did I need a vendor to reaffirm in sentiment analysis what the customer already is telling me? Instead you should focus on insights or affirmations, and use appropriate visualizations to tell the narrative. Unlike the vendor, pick and choose methods and visuals that account for the entire business problem. The vendor ignored the first and more simple question. Coincidentally, during this chapter you will make one of the vendor’s visuals without spending the quarter million dollars they asked for.

3. Tracking sentiment over time

In this section you will track sentiment over time. This type of visualization is often employed in marketing. Let’s say your brand has had a zero point five polarity score month to month in Twitter mentions. You decide to start marketing on Twitter with funny sponsored tweets and hope your brand is perceived as more engaging from the funny ads. As the ad campaign unfolds you can track the sentiment to see the ad’s impact on your brands perception within twitter. Here you are chronologically scoring a book to see if it has a happy or sad ending.

4. Simple frequency analysis

Next you create a ggplot to see which words are most frequently positive or negative. In many of my text mining projects a simple frequency analysis can aid my outcomes. So I like to make frequency analysis part of my EDA, exploratory data analysis, when doing a text mining project. Often I make a visual like you will in this section to help me consume the information quickly. In this case, a sentiment based frequency visual helps to understand the lexical diversity of the polarized text. For example if you were a product manager for an app and the reviews repeatedly mentioned “strange” amid other negative words you may want to review the user interface because people may be finding it strange or unintuitive. In this section’s exercise your visual is based on Moby Dick, generally considered a tragedy. In this case, the visual merely demonstrates Herman Melville, the author’s, polarized word diversity. Nonetheless, the visual can be constructed using any text inner joined with the bing lexicon and you should explore using it on your own text.

5. Let's practice!

Create Your Free Account

By continuing, you accept our Terms of Use, our Privacy Policy and that your data is stored in the USA.