A dataset of four Decades of Italian TV Commercials: visual, audio, and linguistic feature descriptors

We present a feature-level dataset of Italian television commercials spanning over four decades (1980-2024) and including more than 10,000 adverts. Unlike existing corpora, which are often limited in scope or duration, this resource captures the longitudinal evolution of a national advertising ecosystem through structured descriptors rather than raw media. Each commercial is enriched with metadata (title, brand, product category, year…) and segmented into scenes with representative frames from which we extract standardised colour palettes. We analyse audio tracks to extract spectral, chromatic, and rhythmic features, while spoken content is processed into linguistic annotations and psychometric markers. This integrated collection of visual, auditory, and textual descriptors enables cross-modal and temporal studies of advertising strategies and cultural trends. By releasing the dataset openly together with the complete preprocessing code, we ensure transparency and reproducibility, while also providing a methodology that can be readily applied to expand this collection or to build similar resources in other cultural and linguistic contexts, thereby supporting research in communication studies, multimedia analytics, and computational social science.