Sentiment Rings

Sentiment and topic analysis of user reviews for business software.

Management was curious what reviewers were saying on G2 Crowd, a user review aggregator of business software products. I web scraped the reviews using Import.io, then used R and manual inspection to identify sentiment, top words and topics. I created a visual one-pager summarizing my findings that was shared with executives, resulting in social media and review monitoring.

Tools

  • Excel

  • Import.io

  • R

  • Tableau

  • Illustrator

My Role

  • Data Prep

  • Analysis

  • Development

  • Design

Class project & work assignment

Visual Storytelling, Information and Data Visualization, Maryland Institute College of Art (MICA).

Data

G2 Crowd review website. 84 user reviews, with likes, dislikes, recommendation free form text, + other fields. 2016

Type

Infographic


My Process

The Challenge

Craft a visual narrative. Establish a visual hierarchy. Identify sentiment and topics from the review text. Learn R.

Data Prep

Scraped the reviews with Import.io and exported to Excel.

Analysis & Design

I wanted to learn  R and the TidyText package. I ran a sentiment analysis against the BING lexicon, flagging words as positive or negative to understand the user’s emotional intent.

The word count helped me get a general sense of what people were talking about.  Finally I manually read the top words in context to identify topics and comment clusters and tagged each review as 1 (positive) or -1 (negative). Given more time I would have use the LDA package to identify word clusters.

Final Design

I used Tableau to generate the circle sizes, depicting the topics discussed. Dividing each circle by positive and negative sentiment percentage condensed the visualization, essentially combining the packed circles and bar charts into circles.

The charts were exported and manipulated in Illustrator to create the rings and add the story elements. The headline and lead-in explain the purpose and key take-aways. Annotations callout the sentiment percentage and subtopics for the top topics.

Right panel single bar chart and vertical word cloud convey overall sentiment showing the associated top-10 negative and positive words.

The Result

This visualization was shared with executives, resulting in social media and review monitoring.

Previous
Previous

Salary Paths

Next
Next

Concussion Explorer