This study proposes a method to measure consumer lifestyle consciousness through from online reviews in the context of purchases by analyzing online reviews, focusing on wristwatches. Data on 100,248 watches and corresponding 929,240 reviews were extracted from the Amazon Reviews 2023 dataset. From multi-country questionnaires, we derived seed words for nine lifestyle consciousness factors using text preprocessing, synonym expansion, and SBERT-based relevance scoring. Next, a 200-dimensional Word2Vec model was trained on the watch-review corpus and a centroid vector was computed for each consciousness factor by aggregating its seed-word embeddings based on Distributed Dictionary Representation (DDR). We then calculated cosine similarities between centroids to examine contextual similarities among factors and found multiple positive similarity pairs, implying that these consciousness concepts may share review contexts rather than being fully independent. This approach enables marketers and designers to capture multidimensional consumer psychological tendencies from e-commerce data without relying on questionnaire surveys, providing actionable insights for developing marketing strategies.
MAO et al. (Thu,) studied this question.