Recommender System Based on Unstructured Data

Abstract

As the time constraints for completing customer satisfaction surveys are becoming tighter, there emerges a need for developing a new format of surveying customers. The idea is to limit the number of score benchmark questions, and let customers express their opinions in a free format. As a result the collected data will mainly contain open-ended text comments. This chapter presents a strategy to modify the existing recommender system built based on both numerical (structured) and text data (unstructured) to work on unstructured data only to make it work with the new survey format.