Extracting Customer-Related Information for Need Identification

Abstract

The increasing amounts of customer-generated content regarding a product or service published in Social Media are an important source of information for companies. Especially for product development projects or the design of service offers, the unbiased feedback expressed in so-called product reviews is most valuable. However, for the effective use of product review content, the development of automated text processing tools is essential; manual text processing approaches are very time-consuming and thus compromise the benefits provided from the extracted information. To date, automated text mining tools focus the analysis of customers preferences and emotions articulated within a product review. An automated extraction and analysis of customer-related content has not yet been investigated in detail. Customer-related content refers to information within a review, which does not primarily concern the product, but provide information about the customer himself, his usage behavior, personal environment and habits. This information is most generally expressed in an objective manner by the author (i.e. customer) and provides an authentic starting point for the identification of customer needs. Particularly for innovative product development, the consideration of customer habits and personal environment is highly relevant for the derivation of underlying needs, which can be more important than the knowledge of specific preferences regarding a product. The objective of this research is the development and validation of a text mining process for the extraction of objective content from product reviews. To this end, German reviews from Amazon.de regarding two product categories are collected and firstly annotated manually for validation reference. Thereafter, a text mining process is developed comprising text preparation, transformation, classification and performance evaluation. Three different classifiers are applied for performance comparison.