A Cascaded Wearable Product Feature Tag Extraction Method Based on Sentiment Prior

Nowadays wearable products have a large application space in many places. Extracting product feature tags from user comments and mining product advantages and disadvantages found by users during use can effectively improve the design of wearable products and then promote product sales. The current methods can extract tags from sentences through methods like keyword extraction, semantic similarity or multi-label classification. However, these methods have problems such as inaccurate tag extraction and a need of large amounts of labeled data. To address these issues, this paper proposes a hierarchical feature tag extraction algorithm based on aspect-level sentiment prior. Specifically, the user comments are first with aspect-level sentiment recognition to extract opinion words as prior information, and then the semantic similarity model is applied to get the final feature tags. The experimental results show that the proposed method improves the accuracy of product feature tag extraction from user comments in the field of wearable products.