Novel context-based text analytics methods with applications on customer reviews for products and services

Abstract

In chapter 1, we propose a sentence-based deep neural language model that is 1) able to detect customer satisfaction- and dissatisfaction-related elements on online reviews from hotels and restaurants, and 2) capable of providing a list of interpretable sentences with their scores of closeness to customer satisfaction that summarizes a large number of reviews, as well as provides insights to business owners and managers for potentially better understanding of customer opinions. Our model is not only used for text document classification, but it is also used to guide decision-makers with deeper and insightful information from customer feedback. The meaning of the words and short n-grams are context dependent, and it must be understood in the sentence that they appear. We use sentences as the basic elements or features for describing text that helps to understand the meaning of words or n-grams in their context.

In chapter 2, we present a network-based framework to identify and cluster phrases or phrases with context clues in classes of text data using prevalence scores of n-gram structures and their connections. Our framework extends the bag-of-words models by network-based clustering methods to create sub-graphs of connected n-grams for finding context clues. The paths in these sub-graphs represent sequences of words, which form connected phrases with richer contextual meaning. We use our method to identify variations of these phrases and apply the proposed framework to study a collection of customer reviews from TripAdvisor and Yelp. We compare electiveness of our method to standard bag-of-words models such as latent Dirichlet allocation (LDA) and sentiment models.

In chapter 3, we adopt the information asymmetry model from information economics to provide an overview of the interaction between a seller, a buyer, and a reviewer as a system in e-commerce to incorporate not only adverse selection but also moral hazard to explain how online reviews serve as screening and incentive mechanisms to reduce information asymmetry in e-commerce. We argue that online reviews not only act as a screening mechanism to reduce information asymmetry between online sellers and buyers as shown in previous literature, but they also serve as an incentive mechanism from reviewers to mitigate moral hazard incurred by product or service providers. We contribute to information systems and information economics literature by providing an overview of an online review system by adopting the information asymmetry model to the context of online reviews in e-commerce to explain the interaction of sellers, buyers, and reviewers using online reviews as a screening mechanism to mitigate adverse selection and also as an incentive mechanism to reduce moral hazard.