Online shopping platforms often highlight reviews to aid consumers’ decision-making process. The current research proposes that highlighted review should match between the reviewers’ and the browsing consumers’ purchasing goals (profiles). Using Latent Dirichlet Allocation (LDA), an unsupervised machine learning method for topic modeling, we uncovered the hidden profiles that show a reviewer’s original purchasing goal, whether utility-oriented or hedonic-oriented. Subsequent analysis revealed that utility- and hedonic-oriented reviewers differ in certain review-writing and rating behaviors. The paper contributes to the literature by suggesting a new way to understand reviewers’ profiles from text data and resulting review behaviors. We also make a practical recommendation for shopping platforms in highlighting more relevant reviews.