Boosting Dual Quality detection with AI-based social media analysis

Dual Quality (DQ) is the illegal practice of selling products in different countries under the same brand and with identical packaging but different composition or properties. Early detection of such practices poses a significant challenge for competition authorities, exacerbated by the lack of adequate automatic tools. To fill this gap, we propose a novel approach that focuses on identifying dual quality mentions (DQMs) in consumer opinions, which can serve as important indicators of DQ practices. By analyzing consumer opinions collected from online sources, we show that despite the scarcity of DQMs in the available data, they provide valuable insights for competition regulators. Our methodology involves the manual annotation of DQM datasets in three languages (English, German, Polish), followed by the development and training of transformer-based DQM detectors. These detectors exhibit high classification performance, as evidenced by their F1 scores, and thus offer promising avenues for effective support to competition regulators.