Geospatial Sentiment Analysis

The success of a company is often dependent on the quality of their Customer Relationship Management (CRM). Knowledge about customer’s concerns and needs can be a huge advantage over competitors but is hard to gain. Large amounts of textual feedback from customers via surveys or emails has to be manually processed, condensed, and lead to decision makers. As this process is quite expensive and error-prone, CRM data is in practice often neglected. We therefore propose an automatic analysis and visualization approach helping analysts in finding interesting patterns. We combine opinion mining with the geospatial location of a review to enable a context-aware analysis of the CRM data. Instead of overwhelming the user by showing the details first, we visually group similar patterns together and aggregate them by applying Self-Organizing Maps in an interactive analysis application. We extend this approach by integrating temporal and seasonal analyses showing these influences on the CRM data. Our technique is able to cope with unstructured customer feedback data and shows location dependencies of significant terms and sentiments. The capabilities of our approach are shown in a case-study using real-world customer feedback data exploring and describing interesting findings.