Mining Brand Image through the analysis of unstructured online customer reviews: A Machine Learning Approach

One of the most important possessions of a company is the brand image, which plays a fundamental role in the marketing strategy. However, there is not a clear consensus on its definition and much less in its measurement. Online customer reviews are growing exponentially, providing a new source to shape brand image and offering the opportunity for companies to understand their brand perception from customers without contacting them directly. This study applies a combination of brand analysis and sentiment analysis, following by text mining, which compares two different topic modelling approaches, and ends with the evaluation of brand image changes over time. Improved results were obtained restricting the corpus to only adjectives and nouns in their lemmatised form. These results were consistent with previous findings regarding specific attributes that characterise each brand (Apple, HTC, and Samsung). The final proposed framework provides companies an alternative baseline to traditional methods to explore, measure and track brand image in an automated and dynamic way. It also offers a foundation for future advances in understanding customer’s brand perception. Moreover, to the best of our knowledge, the approach proposed in this dissertation is the first to mine brand image using unstructured online customer reviews by applying a machine learning approach. key words: Brand Image (BI), Word-of-Mouth (WoM), Electronic-Word-of-Mouth (e- WoM), Consumer Survey Data (CSD), Consumer-Generated Content (CGC), Latent Dirichlet Allocation (LDA