Retail client wanted to analyze customer comments using text analytics. Customer calls into the Service Center are captured and saved in a raw text format. Y&L analysts extracted Service Center data and derived sentiment utilizing latent Dirichlet allocation, text pooling, and clustering algorithms in R. This data was first explored via Word Clouds from clusters in R-Shiny to understand the data output foundation and strength of the applied sentiment dictionary – positively/negatively weighing emotion based upon idiolect.
An R-Shiny dashboard displayed products and locations broken down by sentiment. Visualizations included Word Clouds sized by occurrence identifying where hot topics were located as well as a neural network for tracing feedback to specific topics.
Tools Used: IBM BigInsight, R, Natural Language Processing, R-Shiny