Latent customer needs elicitation for big-data analysis of online product reviews

Traditional customer needs elicitation methods are often time and cost consuming due to the linguistic analysis of customer needs. Furthermore, many of them are unable to identify latent customer needs, such as interviews and focus groups. This paper proposes a new paradigm of customer needs elicitation based on sentiment analysis of individual product attributes of online product reviews. Support vector machines are used to build prediction models built on the features extracted from a list of affective lexicons based on affective norms for English words and WordNet. The proposed method is able to compile sentiment information on individual product attributes. Such information greatly facilitates the process to elicit customer needs, especially latent ones. We also present a case study to show the potential and feasibility of the proposed method.