User:Ukailin/Mis510proj

is a project for students enrolled in MIS510-Web Mining at the MIS Department, University of Arizona.

My project is about sentiment analysis on Amazon customer reviews. Among various linguistic, statistical, and machine learning techniques in text analysis, sentiment analysis and opinion mining are particularly relevant and useful in marketing research. In the context of marketing, how people feel about an object (either a product, a brand, or a company) plays an important role in their appraisals of the object, which could lead to corresponding behaviors, such as buying or not. For example, studies have been investigating the use of emotional adjectives in twitter messages, and relate those features to brand knowledge (awareness and image), brand relationship (satisfaction, trust, and attachment), and behavioral outcomes (current and future purchases). However, little has been done in decompose (and, later, aggregate) sentiments toward different facets of a target object. A fine granularity sentiment analysis on attribute level, instead of object level, can also be of great value in the marketing decision making. Based on such inputs, managers can adjust marketing actions and focus on certain attributes. Therefore, in my study I hypothesize that when the attribute-based sentiments are aggregated into a single score, the score will be more accurate than a sentiment score from traditional object-based methods.