The traditional product design process begins with the identification of user needs (Ulrich and Eppinger 2008). Traditional methods for needs identification include focus groups, surveys, interviews, and anthropological studies. In this paper, we propose to augment traditional methods for identifying user needs by automatically analyzing user-generated online product reviews. Specifically, we present a supervised, machine learning approach for sentential-level adaptive text extraction and mining. Based upon a set of 9700+ digital camera product reviews gathered in January 2008, we evaluate the approach in three ways. First, we report precision and recall using n-fold cross-validation on labeled data. Second, we compare the recall of automated learning with respect to traditional measures for identifying users and their respective needs. Third, we use multi-dimensional scaling (MDS) to visualize the competitive landscape by mapping existing products in terms of the user needs that they address.
Lee, Thomas Y., "Automatically Learning User Needs from Online Reviews for New Product Design" (2009). AMCIS 2009 Proceedings. 22.