Natural Language Processing for Fashion Trends Detection

Being in a fast-changing and highly competitive environment, fashion companies must constantly adapt to the changing tastes and needs of their consumers. In fact, fashion market trends are among the most important factors influencing consumers’ tastes. Therefore, this paper proposes a fashion trends detection system based on an experts’ approach. Instead of manually scouring tons of fashion magazines and weblogs, this system gathers web text data from such sources and applies text mining tools to detect the trendiest fashion topics discussed in a certain period of time. By defining a list of French fashion-related words and categorizing them into different fashion design elements - namely a garment’s type, color, pattern, material, and style - we came up with a French fashion dictionary used, along with Natural Language Processing techniques, to extract fashion-related words and expressions from the collected blogs. By analyzing the frequency of occurrence and co-occurrence of the extracted words, the proposed framework was able to detect and visualize trends, as well as identify trends’ evolution over time as a first step toward trends prediction.