GETUSP: Goal-oriented topic modeling framework for unique selling point discovery

The Unique Selling Point (USP) is essential for articulating a product's unique strengths to customers, facilitating its competitive success in the market. However, traditional USP discovery approaches often rely on subjective judgment. This limitation leads to a gap between the USP planned by companies and the USP that truly resonates with consumers. To address this gap, topic modeling has been applied to customer reviews, but existing methods face challenges in performing USP-centered clustering, and the derivation of effective USPs is constrained due to ambiguous topic interpretation. We propose GETUSP, a two-stage USP discovery framework designed to extract and recommend USPs from customer reviews. In the first stage, Goal-Oriented Topic Modeling clusters reviews to identify interpretable, sentence-level USP candidates, overcoming the limitations of traditional methods. In the second stage, the Unique Selling Score quantitatively evaluates these candidates by incorporating consumer-perceived benefits and differentiation from competitors. This approach enables effective identification, evaluation, and prioritization of USPs that align with consumer needs and stand out in the market. Experimental results demonstrate that GETUSP outperforms baseline methods in terms of Precision@k and delivers consistent performance across diverse products. These findings confirm its practical applicability and highlight its potential for guiding data-driven USP discovery in various industries.